linux-kernel - Re: [PATCH -next v6 5/5] md: protect md

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <720261bf-2dfd-1298-4664-853a8e5cf132@huaweicloud.com>
Date:   Fri, 14 Apr 2023 09:26:35 +0800
From:   Yu Kuai <yukuai1@...weicloud.com>
To:     Logan Gunthorpe <logang@...tatee.com>,
        Yu Kuai <yukuai1@...weicloud.com>, dm-devel@...hat.com,
        song@...nel.org
Cc:     linux-kernel@...r.kernel.org, linux-raid@...r.kernel.org,
        yi.zhang@...wei.com, yangerkun@...wei.com,
        "yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [PATCH -next v6 5/5] md: protect md_thread with rcu

Hi,

在 2023/04/13 23:10, Logan Gunthorpe 写道:
> 
> 
> On 2023-04-13 05:32, Yu Kuai wrote:
>> From: Yu Kuai <yukuai3@...wei.com>
>>
>> Our test reports a uaf for 'mddev->sync_thread':
>>
>> T1                      T2
>> md_start_sync
>>   md_register_thread
>>   // mddev->sync_thread is set
>> 			raid1d
>> 			 md_check_recovery
>> 			  md_reap_sync_thread
>> 			   md_unregister_thread
>> 			    kfree
>>
>>   md_wakeup_thread
>>    wake_up
>>    ->sync_thread was freed
>>
>> Root cause is that there is a small windown between register thread and
>> wake up thread, where the thread can be freed concurrently.
>>
>> Currently, a global spinlock 'pers_lock' is borrowed to protect
>> 'mddev->thread', this problem can be fixed likewise, however, there are
>> similar problems elsewhere, and use a global lock for all the cases is
>> not good.
>>
>> This patch protect all md_thread with rcu.
>>
>> Signed-off-by: Yu Kuai <yukuai3@...wei.com>
>> ---
>>   drivers/md/md-bitmap.c    | 10 ++++--
>>   drivers/md/md-cluster.c   |  7 ++--
>>   drivers/md/md-multipath.c |  4 +--
>>   drivers/md/md.c           | 69 ++++++++++++++++++---------------------
>>   drivers/md/md.h           |  8 ++---
>>   drivers/md/raid1.c        |  7 ++--
>>   drivers/md/raid1.h        |  2 +-
>>   drivers/md/raid10.c       | 21 +++++++-----
>>   drivers/md/raid10.h       |  2 +-
>>   drivers/md/raid5-cache.c  | 22 ++++++++-----
>>   drivers/md/raid5.c        | 15 +++++----
>>   drivers/md/raid5.h        |  2 +-
>>   12 files changed, 91 insertions(+), 78 deletions(-)
>>
>> diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c
>> index 29fd41ef55a6..ab27f66dbb1f 100644
>> --- a/drivers/md/md-bitmap.c
>> +++ b/drivers/md/md-bitmap.c
>> @@ -1221,13 +1221,19 @@ static bitmap_counter_t *md_bitmap_get_counter(struct bitmap_counts *bitmap,
>>   static void mddev_set_timeout(struct mddev *mddev, unsigned long timeout,
>>   			      bool force)
>>   {
>> -	struct md_thread *thread = mddev->thread;
>> +	struct md_thread *thread;
>> +
>> +	rcu_read_lock();
>> +	thread = rcu_dereference(mddev->thread);
>>   
>>   	if (!thread)
>> -		return;
>> +		goto out;
>>   
>>   	if (force || thread->timeout < MAX_SCHEDULE_TIMEOUT)
>>   		thread->timeout = timeout;
>> +
>> +out:
>> +	rcu_read_unlock();
>>   }
>>   
>>   /*
>> diff --git a/drivers/md/md-cluster.c b/drivers/md/md-cluster.c
>> index 10e0c5381d01..672dfa6a40ec 100644
>> --- a/drivers/md/md-cluster.c
>> +++ b/drivers/md/md-cluster.c
>> @@ -362,8 +362,8 @@ static void __recover_slot(struct mddev *mddev, int slot)
>>   
>>   	set_bit(slot, &cinfo->recovery_map);
>>   	if (!cinfo->recovery_thread) {
>> -		cinfo->recovery_thread = md_register_thread(recover_bitmaps,
>> -				mddev, "recover");
>> +		rcu_assign_pointer(cinfo->recovery_thread,
>> +			md_register_thread(recover_bitmaps, mddev, "recover"));
>>   		if (!cinfo->recovery_thread) {
>>   			pr_warn("md-cluster: Could not create recovery thread\n");
>>   			return;
>> @@ -889,7 +889,8 @@ static int join(struct mddev *mddev, int nodes)
>>   	}
>>   	/* Initiate the communication resources */
>>   	ret = -ENOMEM;
>> -	cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv");
>> +	rcu_assign_pointer(cinfo->recv_thread,
>> +			md_register_thread(recv_daemon, mddev, "cluster_recv"));
>>   	if (!cinfo->recv_thread) {
> 
> This looks like it'll hit sparse warnings. without an
> rcu_access_pointer(). Might be nicer to use a temporary variable, check
> that it's not null, then call rcu_assign_pointer().

Sorry this is because I missed following change:

diff --git a/drivers/md/md-cluster.c b/drivers/md/md-cluster.c
index 672dfa6a40ec..2f64673b6ed2 100644
--- a/drivers/md/md-cluster.c
+++ b/drivers/md/md-cluster.c
@@ -75,14 +75,14 @@ struct md_cluster_info {
         sector_t suspend_hi;
         int suspend_from; /* the slot which broadcast suspend_lo/hi */

-       struct md_thread *recovery_thread;
+       struct md_thread __rcu *recovery_thread;
         unsigned long recovery_map;
         /* communication loc resources */
         struct dlm_lock_resource *ack_lockres;
         struct dlm_lock_resource *message_lockres;
         struct dlm_lock_resource *token_lockres;
         struct dlm_lock_resource *no_new_dev_lockres;
-       struct md_thread *recv_thread;
+       struct md_thread __rcu *recv_thread;
         struct completion newdisk_completion;
         wait_queue_head_t wait;
         unsigned long state;

Thanks,
Kuai