Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH 3/3] ext4: add rcu to prevent use after free when umount
 filesystem
To:     Jan Kara <jack@suse.cz>
CC:     <linux-ext4@vger.kernel.org>, <tytso@mit.edu>,
        <adilger.kernel@dilger.ca>, <yukuai3@huawei.com>
References: <20210408113618.1033785-1-yi.zhang@huawei.com>
 <20210408113618.1033785-4-yi.zhang@huawei.com>
 <20210408135603.GD3271@quack2.suse.cz>
From:   Zhang Yi <yi.zhang@huawei.com>
Message-ID: <d75b2ebd-33fc-8f1e-a3b4-d4715ef85314@huawei.com>
Date:   Thu, 8 Apr 2021 22:38:08 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Thunderbird/78.3.1
MIME-Version: 1.0
In-Reply-To: <20210408135603.GD3271@quack2.suse.cz>
Content-Type: text/plain; charset="utf-8"
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2021/4/8 21:56, Jan Kara wrote:
> On Thu 08-04-21 19:36:18, Zhang Yi wrote:
>> There is a race between bdev_try_to_free_page() and
>> jbd2_journal_destroy() that could end up triggering a use after free
>> issue about journal.
>>
>> drop cache                           umount filesystem
>> bdev_try_to_free_page()
>>  get journal
>>  jbd2_journal_try_to_free_buffers()  ext4_put_super()
>>                                       kfree(journal)
>>    access journal <-- lead to UAF
>>
>> The above race also could happens between the bdev_try_to_free_page()
>> and the error path of ext4_fill_super(). This patch avoid this race by
>> add rcu protection around accessing sbi->s_journal in
>> bdev_try_to_free_page() and destroy the journal after an rcu grace
>> period.
>>
>> Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
> 
> OK, I see the problem. But cannot the use-after-free happen even for the
> superblock itself (e.g., EXT4_SB(sb)->s_journal dereference)? I don't see
> anything preventing that as blkdev_releasepage() just shamelessly does:
> 
> super = BDEV_I(page->mapping->host)->bdev.bd_super
> 
Hi, Jan.

I checked the superblock. In theory, the bdev_try_to_free_page() is invoked
with page locked, the umount process will wait the page unlock on
kill_block_super()->..->kill_bdev()->truncate_inode_pages_range() before free
superblock, so I guess the use-after-free problem couldn't happen in general.
But I think it's fragile and may invalidate if the bdev has more than one
operners(__blkdev_put() call kill_bdev only if bd_openers becomes zero)?
I will check it.

Thanks,
Yi.

> without making sure the sb cannot go away the instant we load a pointer to
> it. Or am I missing something Ted? If I'm right, we'd need some careful
> sprinkling of RCU, READ_ONCE(), and careful superblock reference grabbing
> to make bdev_try_to_free_page() safe against concurrent
> kill_block_super()...
> 
> 								Honza
> 
>> ---
>>  fs/ext4/super.c      | 33 ++++++++++++++++++++++++---------
>>  fs/jbd2/journal.c    | 30 +++++++++++++++++++++++++++---
>>  include/linux/jbd2.h | 11 ++++++++++-
>>  3 files changed, 61 insertions(+), 13 deletions(-)
>>
>> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
>> index 02ba47a5bc70..6bbaadc5357b 100644
>> --- a/fs/ext4/super.c
>> +++ b/fs/ext4/super.c
>> @@ -1150,6 +1150,21 @@ static inline void ext4_quota_off_umount(struct super_block *sb)
>>  }
>>  #endif
>>  
>> +static int ext4_journal_release(struct ext4_sb_info *sbi)
>> +{
>> +	journal_t *journal = sbi->s_journal;
>> +	int ret;
>> +
>> +	ret = jbd2_journal_release(journal);
>> +	sbi->s_journal = NULL;
>> +	/*
>> +	 * Call rcu to prevent racing with bdev_try_to_free_page()
>> +	 * accessing the journal at the same time.
>> +	 */
>> +	call_rcu(&journal->j_rcu, jbd2_journal_release_rcu);
>> +	return ret;
>> +}
>> +
>>  static void ext4_put_super(struct super_block *sb)
>>  {
>>  	struct ext4_sb_info *sbi = EXT4_SB(sb);
>> @@ -1174,11 +1189,9 @@ static void ext4_put_super(struct super_block *sb)
>>  
>>  	if (sbi->s_journal) {
>>  		aborted = is_journal_aborted(sbi->s_journal);
>> -		err = jbd2_journal_destroy(sbi->s_journal);
>> -		sbi->s_journal = NULL;
>> -		if ((err < 0) && !aborted) {
>> +		err = ext4_journal_release(sbi);
>> +		if ((err < 0) && !aborted)
>>  			ext4_abort(sb, -err, "Couldn't clean up the journal");
>> -		}
>>  	}
>>  
>>  	ext4_es_unregister_shrinker(sbi);
>> @@ -1449,14 +1462,18 @@ static int ext4_nfs_commit_metadata(struct inode *inode)
>>  static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
>>  				 gfp_t wait)
>>  {
>> -	journal_t *journal = EXT4_SB(sb)->s_journal;
>> +	journal_t *journal;
>>  	int ret = 0;
>>  
>>  	WARN_ON(PageChecked(page));
>>  	if (!page_has_buffers(page))
>>  		return 0;
>> +
>> +	rcu_read_lock();
>> +	journal = READ_ONCE(EXT4_SB(sb)->s_journal);
>>  	if (journal)
>>  		ret = jbd2_journal_try_to_free_buffers(journal, page);
>> +	rcu_read_unlock();
>>  	if (!ret)
>>  		return try_to_free_buffers(page);
>>  	return 0;
>> @@ -5146,10 +5163,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>>  	ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
>>  	sbi->s_ea_block_cache = NULL;
>>  
>> -	if (sbi->s_journal) {
>> -		jbd2_journal_destroy(sbi->s_journal);
>> -		sbi->s_journal = NULL;
>> -	}
>> +	if (sbi->s_journal)
>> +		ext4_journal_release(sbi);
>>  failed_mount3a:
>>  	ext4_es_unregister_shrinker(sbi);
>>  failed_mount3:
>> diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
>> index 2dc944442802..071caaaa9de1 100644
>> --- a/fs/jbd2/journal.c
>> +++ b/fs/jbd2/journal.c
>> @@ -76,6 +76,8 @@ EXPORT_SYMBOL(jbd2_journal_check_available_features);
>>  EXPORT_SYMBOL(jbd2_journal_set_features);
>>  EXPORT_SYMBOL(jbd2_journal_load);
>>  EXPORT_SYMBOL(jbd2_journal_destroy);
>> +EXPORT_SYMBOL(jbd2_journal_release);
>> +EXPORT_SYMBOL(jbd2_journal_release_rcu);
>>  EXPORT_SYMBOL(jbd2_journal_abort);
>>  EXPORT_SYMBOL(jbd2_journal_errno);
>>  EXPORT_SYMBOL(jbd2_journal_ack_err);
>> @@ -1951,14 +1953,14 @@ int jbd2_journal_load(journal_t *journal)
>>  }
>>  
>>  /**
>> - * jbd2_journal_destroy() - Release a journal_t structure.
>> + * jbd2_journal_release() - Release a journal_t structure.
>>   * @journal: Journal to act on.
>>   *
>>   * Release a journal_t structure once it is no longer in use by the
>>   * journaled object.
>>   * Return <0 if we couldn't clean up the journal.
>>   */
>> -int jbd2_journal_destroy(journal_t *journal)
>> +int jbd2_journal_release(journal_t *journal)
>>  {
>>  	int err = 0;
>>  
>> @@ -2021,11 +2023,33 @@ int jbd2_journal_destroy(journal_t *journal)
>>  		crypto_free_shash(journal->j_chksum_driver);
>>  	kfree(journal->j_fc_wbuf);
>>  	kfree(journal->j_wbuf);
>> -	kfree(journal);
>>  
>>  	return err;
>>  }
>>  
>> +/**
>> + * jbd2_journal_release_rcu() - Free a journal_t structure.
>> + * @rcu: rcu list node relate to the journal want to free.
>> + *
>> + * Freeing a journal_t structure after a rcu grace period.
>> + */
>> +void jbd2_journal_release_rcu(struct rcu_head *rcu)
>> +{
>> +	kfree(container_of(rcu, journal_t, j_rcu));
>> +}
>> +
>> +/**
>> + * jbd2_journal_destroy() - Release and free a journal_t structure.
>> + * @journal: Journal to act on.
>> + *
>> + * Release and free a journal_t structure once it is no longer in use
>> + * by the journaled object.
>> + */
>> +void jbd2_journal_destroy(journal_t *journal)
>> +{
>> +	jbd2_journal_release(journal);
>> +	kfree(journal);
>> +}
>>  
>>  /**
>>   * jbd2_journal_check_used_features() - Check if features specified are used.
>> diff --git a/include/linux/jbd2.h b/include/linux/jbd2.h
>> index 99d3cd051ac3..39a8d04596a2 100644
>> --- a/include/linux/jbd2.h
>> +++ b/include/linux/jbd2.h
>> @@ -1238,6 +1238,13 @@ struct journal_s
>>  	 */
>>  	__u32 j_csum_seed;
>>  
>> +	/**
>> +	 * @j_rcu:
>> +	 *
>> +	 * Prevent racing between accessing and destroy at the same time.
>> +	 */
>> +	struct rcu_head j_rcu;
>> +
>>  #ifdef CONFIG_DEBUG_LOCK_ALLOC
>>  	/**
>>  	 * @j_trans_commit_map:
>> @@ -1509,7 +1516,9 @@ extern int	   jbd2_journal_set_features
>>  extern void	   jbd2_journal_clear_features
>>  		   (journal_t *, unsigned long, unsigned long, unsigned long);
>>  extern int	   jbd2_journal_load       (journal_t *journal);
>> -extern int	   jbd2_journal_destroy    (journal_t *);
>> +extern void	   jbd2_journal_destroy    (journal_t *);
>> +extern int	   jbd2_journal_release    (journal_t *);
>> +extern void	   jbd2_journal_release_rcu     (struct rcu_head *rcu);
>>  extern int	   jbd2_journal_recover    (journal_t *journal);
>>  extern int	   jbd2_journal_wipe       (journal_t *, int);
>>  extern int	   jbd2_journal_skip_recovery	(journal_t *);
>> -- 
>> 2.25.4
>>