Revision 462a8e08e0e6287e5ce13187257edbf24213ed03 authored by David Chen on 09 February 2023, 17:48:28 UTC, committed by Linus Torvalds on 12 February 2023, 18:30:05 UTC
When we upgraded our kernel, we started seeing some page corruption like
the following consistently:

  BUG: Bad page state in process ganesha.nfsd  pfn:1304ca
  page:0000000022261c55 refcount:0 mapcount:-128 mapping:0000000000000000 index:0x0 pfn:0x1304ca
  flags: 0x17ffffc0000000()
  raw: 0017ffffc0000000 ffff8a513ffd4c98 ffffeee24b35ec08 0000000000000000
  raw: 0000000000000000 0000000000000001 00000000ffffff7f 0000000000000000
  page dumped because: nonzero mapcount
  CPU: 0 PID: 15567 Comm: ganesha.nfsd Kdump: loaded Tainted: P    B      O      5.10.158-1.nutanix.20221209.el7.x86_64 #1
  Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 04/05/2016
  Call Trace:
   dump_stack+0x74/0x96
   bad_page.cold+0x63/0x94
   check_new_page_bad+0x6d/0x80
   rmqueue+0x46e/0x970
   get_page_from_freelist+0xcb/0x3f0
   ? _cond_resched+0x19/0x40
   __alloc_pages_nodemask+0x164/0x300
   alloc_pages_current+0x87/0xf0
   skb_page_frag_refill+0x84/0x110
   ...

Sometimes, it would also show up as corruption in the free list pointer
and cause crashes.

After bisecting the issue, we found the issue started from commit
e320d3012d25 ("mm/page_alloc.c: fix freeing non-compound pages"):

	if (put_page_testzero(page))
		free_the_page(page, order);
	else if (!PageHead(page))
		while (order-- > 0)
			free_the_page(page + (1 << order), order);

So the problem is the check PageHead is racy because at this point we
already dropped our reference to the page.  So even if we came in with
compound page, the page can already be freed and PageHead can return
false and we will end up freeing all the tail pages causing double free.

Fixes: e320d3012d25 ("mm/page_alloc.c: fix freeing non-compound pages")
Link: https://lore.kernel.org/lkml/BYAPR02MB448855960A9656EEA81141FC94D99@BYAPR02MB4488.namprd02.prod.outlook.com/
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Signed-off-by: Chunwei Chen <david.chen@nutanix.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent f339c25
Raw File
watch_queue.c
// SPDX-License-Identifier: GPL-2.0
/* Watch queue and general notification mechanism, built on pipes
 *
 * Copyright (C) 2020 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 *
 * See Documentation/core-api/watch_queue.rst
 */

#define pr_fmt(fmt) "watchq: " fmt
#include <linux/module.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/printk.h>
#include <linux/miscdevice.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/poll.h>
#include <linux/uaccess.h>
#include <linux/vmalloc.h>
#include <linux/file.h>
#include <linux/security.h>
#include <linux/cred.h>
#include <linux/sched/signal.h>
#include <linux/watch_queue.h>
#include <linux/pipe_fs_i.h>

MODULE_DESCRIPTION("Watch queue");
MODULE_AUTHOR("Red Hat, Inc.");
MODULE_LICENSE("GPL");

#define WATCH_QUEUE_NOTE_SIZE 128
#define WATCH_QUEUE_NOTES_PER_PAGE (PAGE_SIZE / WATCH_QUEUE_NOTE_SIZE)

/*
 * This must be called under the RCU read-lock, which makes
 * sure that the wqueue still exists. It can then take the lock,
 * and check that the wqueue hasn't been destroyed, which in
 * turn makes sure that the notification pipe still exists.
 */
static inline bool lock_wqueue(struct watch_queue *wqueue)
{
	spin_lock_bh(&wqueue->lock);
	if (unlikely(wqueue->defunct)) {
		spin_unlock_bh(&wqueue->lock);
		return false;
	}
	return true;
}

static inline void unlock_wqueue(struct watch_queue *wqueue)
{
	spin_unlock_bh(&wqueue->lock);
}

static void watch_queue_pipe_buf_release(struct pipe_inode_info *pipe,
					 struct pipe_buffer *buf)
{
	struct watch_queue *wqueue = (struct watch_queue *)buf->private;
	struct page *page;
	unsigned int bit;

	/* We need to work out which note within the page this refers to, but
	 * the note might have been maximum size, so merely ANDing the offset
	 * off doesn't work.  OTOH, the note must've been more than zero size.
	 */
	bit = buf->offset + buf->len;
	if ((bit & (WATCH_QUEUE_NOTE_SIZE - 1)) == 0)
		bit -= WATCH_QUEUE_NOTE_SIZE;
	bit /= WATCH_QUEUE_NOTE_SIZE;

	page = buf->page;
	bit += page->index;

	set_bit(bit, wqueue->notes_bitmap);
	generic_pipe_buf_release(pipe, buf);
}

// No try_steal function => no stealing
#define watch_queue_pipe_buf_try_steal NULL

/* New data written to a pipe may be appended to a buffer with this type. */
static const struct pipe_buf_operations watch_queue_pipe_buf_ops = {
	.release	= watch_queue_pipe_buf_release,
	.try_steal	= watch_queue_pipe_buf_try_steal,
	.get		= generic_pipe_buf_get,
};

/*
 * Post a notification to a watch queue.
 *
 * Must be called with the RCU lock for reading, and the
 * watch_queue lock held, which guarantees that the pipe
 * hasn't been released.
 */
static bool post_one_notification(struct watch_queue *wqueue,
				  struct watch_notification *n)
{
	void *p;
	struct pipe_inode_info *pipe = wqueue->pipe;
	struct pipe_buffer *buf;
	struct page *page;
	unsigned int head, tail, mask, note, offset, len;
	bool done = false;

	if (!pipe)
		return false;

	spin_lock_irq(&pipe->rd_wait.lock);

	mask = pipe->ring_size - 1;
	head = pipe->head;
	tail = pipe->tail;
	if (pipe_full(head, tail, pipe->ring_size))
		goto lost;

	note = find_first_bit(wqueue->notes_bitmap, wqueue->nr_notes);
	if (note >= wqueue->nr_notes)
		goto lost;

	page = wqueue->notes[note / WATCH_QUEUE_NOTES_PER_PAGE];
	offset = note % WATCH_QUEUE_NOTES_PER_PAGE * WATCH_QUEUE_NOTE_SIZE;
	get_page(page);
	len = n->info & WATCH_INFO_LENGTH;
	p = kmap_atomic(page);
	memcpy(p + offset, n, len);
	kunmap_atomic(p);

	buf = &pipe->bufs[head & mask];
	buf->page = page;
	buf->private = (unsigned long)wqueue;
	buf->ops = &watch_queue_pipe_buf_ops;
	buf->offset = offset;
	buf->len = len;
	buf->flags = PIPE_BUF_FLAG_WHOLE;
	smp_store_release(&pipe->head, head + 1); /* vs pipe_read() */

	if (!test_and_clear_bit(note, wqueue->notes_bitmap)) {
		spin_unlock_irq(&pipe->rd_wait.lock);
		BUG();
	}
	wake_up_interruptible_sync_poll_locked(&pipe->rd_wait, EPOLLIN | EPOLLRDNORM);
	done = true;

out:
	spin_unlock_irq(&pipe->rd_wait.lock);
	if (done)
		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
	return done;

lost:
	buf = &pipe->bufs[(head - 1) & mask];
	buf->flags |= PIPE_BUF_FLAG_LOSS;
	goto out;
}

/*
 * Apply filter rules to a notification.
 */
static bool filter_watch_notification(const struct watch_filter *wf,
				      const struct watch_notification *n)
{
	const struct watch_type_filter *wt;
	unsigned int st_bits = sizeof(wt->subtype_filter[0]) * 8;
	unsigned int st_index = n->subtype / st_bits;
	unsigned int st_bit = 1U << (n->subtype % st_bits);
	int i;

	if (!test_bit(n->type, wf->type_filter))
		return false;

	for (i = 0; i < wf->nr_filters; i++) {
		wt = &wf->filters[i];
		if (n->type == wt->type &&
		    (wt->subtype_filter[st_index] & st_bit) &&
		    (n->info & wt->info_mask) == wt->info_filter)
			return true;
	}

	return false; /* If there is a filter, the default is to reject. */
}

/**
 * __post_watch_notification - Post an event notification
 * @wlist: The watch list to post the event to.
 * @n: The notification record to post.
 * @cred: The creds of the process that triggered the notification.
 * @id: The ID to match on the watch.
 *
 * Post a notification of an event into a set of watch queues and let the users
 * know.
 *
 * The size of the notification should be set in n->info & WATCH_INFO_LENGTH and
 * should be in units of sizeof(*n).
 */
void __post_watch_notification(struct watch_list *wlist,
			       struct watch_notification *n,
			       const struct cred *cred,
			       u64 id)
{
	const struct watch_filter *wf;
	struct watch_queue *wqueue;
	struct watch *watch;

	if (((n->info & WATCH_INFO_LENGTH) >> WATCH_INFO_LENGTH__SHIFT) == 0) {
		WARN_ON(1);
		return;
	}

	rcu_read_lock();

	hlist_for_each_entry_rcu(watch, &wlist->watchers, list_node) {
		if (watch->id != id)
			continue;
		n->info &= ~WATCH_INFO_ID;
		n->info |= watch->info_id;

		wqueue = rcu_dereference(watch->queue);
		wf = rcu_dereference(wqueue->filter);
		if (wf && !filter_watch_notification(wf, n))
			continue;

		if (security_post_notification(watch->cred, cred, n) < 0)
			continue;

		if (lock_wqueue(wqueue)) {
			post_one_notification(wqueue, n);
			unlock_wqueue(wqueue);
		}
	}

	rcu_read_unlock();
}
EXPORT_SYMBOL(__post_watch_notification);

/*
 * Allocate sufficient pages to preallocation for the requested number of
 * notifications.
 */
long watch_queue_set_size(struct pipe_inode_info *pipe, unsigned int nr_notes)
{
	struct watch_queue *wqueue = pipe->watch_queue;
	struct page **pages;
	unsigned long *bitmap;
	unsigned long user_bufs;
	int ret, i, nr_pages;

	if (!wqueue)
		return -ENODEV;
	if (wqueue->notes)
		return -EBUSY;

	if (nr_notes < 1 ||
	    nr_notes > 512) /* TODO: choose a better hard limit */
		return -EINVAL;

	nr_pages = (nr_notes + WATCH_QUEUE_NOTES_PER_PAGE - 1);
	nr_pages /= WATCH_QUEUE_NOTES_PER_PAGE;
	user_bufs = account_pipe_buffers(pipe->user, pipe->nr_accounted, nr_pages);

	if (nr_pages > pipe->max_usage &&
	    (too_many_pipe_buffers_hard(user_bufs) ||
	     too_many_pipe_buffers_soft(user_bufs)) &&
	    pipe_is_unprivileged_user()) {
		ret = -EPERM;
		goto error;
	}

	nr_notes = nr_pages * WATCH_QUEUE_NOTES_PER_PAGE;
	ret = pipe_resize_ring(pipe, roundup_pow_of_two(nr_notes));
	if (ret < 0)
		goto error;

	pages = kcalloc(sizeof(struct page *), nr_pages, GFP_KERNEL);
	if (!pages)
		goto error;

	for (i = 0; i < nr_pages; i++) {
		pages[i] = alloc_page(GFP_KERNEL);
		if (!pages[i])
			goto error_p;
		pages[i]->index = i * WATCH_QUEUE_NOTES_PER_PAGE;
	}

	bitmap = bitmap_alloc(nr_notes, GFP_KERNEL);
	if (!bitmap)
		goto error_p;

	bitmap_fill(bitmap, nr_notes);
	wqueue->notes = pages;
	wqueue->notes_bitmap = bitmap;
	wqueue->nr_pages = nr_pages;
	wqueue->nr_notes = nr_notes;
	return 0;

error_p:
	while (--i >= 0)
		__free_page(pages[i]);
	kfree(pages);
error:
	(void) account_pipe_buffers(pipe->user, nr_pages, pipe->nr_accounted);
	return ret;
}

/*
 * Set the filter on a watch queue.
 */
long watch_queue_set_filter(struct pipe_inode_info *pipe,
			    struct watch_notification_filter __user *_filter)
{
	struct watch_notification_type_filter *tf;
	struct watch_notification_filter filter;
	struct watch_type_filter *q;
	struct watch_filter *wfilter;
	struct watch_queue *wqueue = pipe->watch_queue;
	int ret, nr_filter = 0, i;

	if (!wqueue)
		return -ENODEV;

	if (!_filter) {
		/* Remove the old filter */
		wfilter = NULL;
		goto set;
	}

	/* Grab the user's filter specification */
	if (copy_from_user(&filter, _filter, sizeof(filter)) != 0)
		return -EFAULT;
	if (filter.nr_filters == 0 ||
	    filter.nr_filters > 16 ||
	    filter.__reserved != 0)
		return -EINVAL;

	tf = memdup_user(_filter->filters, filter.nr_filters * sizeof(*tf));
	if (IS_ERR(tf))
		return PTR_ERR(tf);

	ret = -EINVAL;
	for (i = 0; i < filter.nr_filters; i++) {
		if ((tf[i].info_filter & ~tf[i].info_mask) ||
		    tf[i].info_mask & WATCH_INFO_LENGTH)
			goto err_filter;
		/* Ignore any unknown types */
		if (tf[i].type >= WATCH_TYPE__NR)
			continue;
		nr_filter++;
	}

	/* Now we need to build the internal filter from only the relevant
	 * user-specified filters.
	 */
	ret = -ENOMEM;
	wfilter = kzalloc(struct_size(wfilter, filters, nr_filter), GFP_KERNEL);
	if (!wfilter)
		goto err_filter;
	wfilter->nr_filters = nr_filter;

	q = wfilter->filters;
	for (i = 0; i < filter.nr_filters; i++) {
		if (tf[i].type >= WATCH_TYPE__NR)
			continue;

		q->type			= tf[i].type;
		q->info_filter		= tf[i].info_filter;
		q->info_mask		= tf[i].info_mask;
		q->subtype_filter[0]	= tf[i].subtype_filter[0];
		__set_bit(q->type, wfilter->type_filter);
		q++;
	}

	kfree(tf);
set:
	pipe_lock(pipe);
	wfilter = rcu_replace_pointer(wqueue->filter, wfilter,
				      lockdep_is_held(&pipe->mutex));
	pipe_unlock(pipe);
	if (wfilter)
		kfree_rcu(wfilter, rcu);
	return 0;

err_filter:
	kfree(tf);
	return ret;
}

static void __put_watch_queue(struct kref *kref)
{
	struct watch_queue *wqueue =
		container_of(kref, struct watch_queue, usage);
	struct watch_filter *wfilter;
	int i;

	for (i = 0; i < wqueue->nr_pages; i++)
		__free_page(wqueue->notes[i]);
	kfree(wqueue->notes);
	bitmap_free(wqueue->notes_bitmap);

	wfilter = rcu_access_pointer(wqueue->filter);
	if (wfilter)
		kfree_rcu(wfilter, rcu);
	kfree_rcu(wqueue, rcu);
}

/**
 * put_watch_queue - Dispose of a ref on a watchqueue.
 * @wqueue: The watch queue to unref.
 */
void put_watch_queue(struct watch_queue *wqueue)
{
	kref_put(&wqueue->usage, __put_watch_queue);
}
EXPORT_SYMBOL(put_watch_queue);

static void free_watch(struct rcu_head *rcu)
{
	struct watch *watch = container_of(rcu, struct watch, rcu);

	put_watch_queue(rcu_access_pointer(watch->queue));
	atomic_dec(&watch->cred->user->nr_watches);
	put_cred(watch->cred);
	kfree(watch);
}

static void __put_watch(struct kref *kref)
{
	struct watch *watch = container_of(kref, struct watch, usage);

	call_rcu(&watch->rcu, free_watch);
}

/*
 * Discard a watch.
 */
static void put_watch(struct watch *watch)
{
	kref_put(&watch->usage, __put_watch);
}

/**
 * init_watch - Initialise a watch
 * @watch: The watch to initialise.
 * @wqueue: The queue to assign.
 *
 * Initialise a watch and set the watch queue.
 */
void init_watch(struct watch *watch, struct watch_queue *wqueue)
{
	kref_init(&watch->usage);
	INIT_HLIST_NODE(&watch->list_node);
	INIT_HLIST_NODE(&watch->queue_node);
	rcu_assign_pointer(watch->queue, wqueue);
}

static int add_one_watch(struct watch *watch, struct watch_list *wlist, struct watch_queue *wqueue)
{
	const struct cred *cred;
	struct watch *w;

	hlist_for_each_entry(w, &wlist->watchers, list_node) {
		struct watch_queue *wq = rcu_access_pointer(w->queue);
		if (wqueue == wq && watch->id == w->id)
			return -EBUSY;
	}

	cred = current_cred();
	if (atomic_inc_return(&cred->user->nr_watches) > task_rlimit(current, RLIMIT_NOFILE)) {
		atomic_dec(&cred->user->nr_watches);
		return -EAGAIN;
	}

	watch->cred = get_cred(cred);
	rcu_assign_pointer(watch->watch_list, wlist);

	kref_get(&wqueue->usage);
	kref_get(&watch->usage);
	hlist_add_head(&watch->queue_node, &wqueue->watches);
	hlist_add_head_rcu(&watch->list_node, &wlist->watchers);
	return 0;
}

/**
 * add_watch_to_object - Add a watch on an object to a watch list
 * @watch: The watch to add
 * @wlist: The watch list to add to
 *
 * @watch->queue must have been set to point to the queue to post notifications
 * to and the watch list of the object to be watched.  @watch->cred must also
 * have been set to the appropriate credentials and a ref taken on them.
 *
 * The caller must pin the queue and the list both and must hold the list
 * locked against racing watch additions/removals.
 */
int add_watch_to_object(struct watch *watch, struct watch_list *wlist)
{
	struct watch_queue *wqueue;
	int ret = -ENOENT;

	rcu_read_lock();

	wqueue = rcu_access_pointer(watch->queue);
	if (lock_wqueue(wqueue)) {
		spin_lock(&wlist->lock);
		ret = add_one_watch(watch, wlist, wqueue);
		spin_unlock(&wlist->lock);
		unlock_wqueue(wqueue);
	}

	rcu_read_unlock();
	return ret;
}
EXPORT_SYMBOL(add_watch_to_object);

/**
 * remove_watch_from_object - Remove a watch or all watches from an object.
 * @wlist: The watch list to remove from
 * @wq: The watch queue of interest (ignored if @all is true)
 * @id: The ID of the watch to remove (ignored if @all is true)
 * @all: True to remove all objects
 *
 * Remove a specific watch or all watches from an object.  A notification is
 * sent to the watcher to tell them that this happened.
 */
int remove_watch_from_object(struct watch_list *wlist, struct watch_queue *wq,
			     u64 id, bool all)
{
	struct watch_notification_removal n;
	struct watch_queue *wqueue;
	struct watch *watch;
	int ret = -EBADSLT;

	rcu_read_lock();

again:
	spin_lock(&wlist->lock);
	hlist_for_each_entry(watch, &wlist->watchers, list_node) {
		if (all ||
		    (watch->id == id && rcu_access_pointer(watch->queue) == wq))
			goto found;
	}
	spin_unlock(&wlist->lock);
	goto out;

found:
	ret = 0;
	hlist_del_init_rcu(&watch->list_node);
	rcu_assign_pointer(watch->watch_list, NULL);
	spin_unlock(&wlist->lock);

	/* We now own the reference on watch that used to belong to wlist. */

	n.watch.type = WATCH_TYPE_META;
	n.watch.subtype = WATCH_META_REMOVAL_NOTIFICATION;
	n.watch.info = watch->info_id | watch_sizeof(n.watch);
	n.id = id;
	if (id != 0)
		n.watch.info = watch->info_id | watch_sizeof(n);

	wqueue = rcu_dereference(watch->queue);

	if (lock_wqueue(wqueue)) {
		post_one_notification(wqueue, &n.watch);

		if (!hlist_unhashed(&watch->queue_node)) {
			hlist_del_init_rcu(&watch->queue_node);
			put_watch(watch);
		}

		unlock_wqueue(wqueue);
	}

	if (wlist->release_watch) {
		void (*release_watch)(struct watch *);

		release_watch = wlist->release_watch;
		rcu_read_unlock();
		(*release_watch)(watch);
		rcu_read_lock();
	}
	put_watch(watch);

	if (all && !hlist_empty(&wlist->watchers))
		goto again;
out:
	rcu_read_unlock();
	return ret;
}
EXPORT_SYMBOL(remove_watch_from_object);

/*
 * Remove all the watches that are contributory to a queue.  This has the
 * potential to race with removal of the watches by the destruction of the
 * objects being watched or with the distribution of notifications.
 */
void watch_queue_clear(struct watch_queue *wqueue)
{
	struct watch_list *wlist;
	struct watch *watch;
	bool release;

	rcu_read_lock();
	spin_lock_bh(&wqueue->lock);

	/* Prevent new notifications from being stored. */
	wqueue->defunct = true;

	while (!hlist_empty(&wqueue->watches)) {
		watch = hlist_entry(wqueue->watches.first, struct watch, queue_node);
		hlist_del_init_rcu(&watch->queue_node);
		/* We now own a ref on the watch. */
		spin_unlock_bh(&wqueue->lock);

		/* We can't do the next bit under the queue lock as we need to
		 * get the list lock - which would cause a deadlock if someone
		 * was removing from the opposite direction at the same time or
		 * posting a notification.
		 */
		wlist = rcu_dereference(watch->watch_list);
		if (wlist) {
			void (*release_watch)(struct watch *);

			spin_lock(&wlist->lock);

			release = !hlist_unhashed(&watch->list_node);
			if (release) {
				hlist_del_init_rcu(&watch->list_node);
				rcu_assign_pointer(watch->watch_list, NULL);

				/* We now own a second ref on the watch. */
			}

			release_watch = wlist->release_watch;
			spin_unlock(&wlist->lock);

			if (release) {
				if (release_watch) {
					rcu_read_unlock();
					/* This might need to call dput(), so
					 * we have to drop all the locks.
					 */
					(*release_watch)(watch);
					rcu_read_lock();
				}
				put_watch(watch);
			}
		}

		put_watch(watch);
		spin_lock_bh(&wqueue->lock);
	}

	spin_unlock_bh(&wqueue->lock);
	rcu_read_unlock();
}

/**
 * get_watch_queue - Get a watch queue from its file descriptor.
 * @fd: The fd to query.
 */
struct watch_queue *get_watch_queue(int fd)
{
	struct pipe_inode_info *pipe;
	struct watch_queue *wqueue = ERR_PTR(-EINVAL);
	struct fd f;

	f = fdget(fd);
	if (f.file) {
		pipe = get_pipe_info(f.file, false);
		if (pipe && pipe->watch_queue) {
			wqueue = pipe->watch_queue;
			kref_get(&wqueue->usage);
		}
		fdput(f);
	}

	return wqueue;
}
EXPORT_SYMBOL(get_watch_queue);

/*
 * Initialise a watch queue
 */
int watch_queue_init(struct pipe_inode_info *pipe)
{
	struct watch_queue *wqueue;

	wqueue = kzalloc(sizeof(*wqueue), GFP_KERNEL);
	if (!wqueue)
		return -ENOMEM;

	wqueue->pipe = pipe;
	kref_init(&wqueue->usage);
	spin_lock_init(&wqueue->lock);
	INIT_HLIST_HEAD(&wqueue->watches);

	pipe->watch_queue = wqueue;
	return 0;
}
back to top