Received-SPF: pass (google.com: domain of linux-kernel+bounces-209852-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Message-ID: <64a8dcab-4f3f-49de-9940-129307371c84@vivo.com>
Date: Tue, 11 Jun 2024 20:28:26 +0800
User-Agent: Mozilla Thunderbird
From: YangYang <yang.yang@vivo.com>
Subject: Re: [PATCH v3] sbitmap: fix io hung due to race on
 sbitmap_word::cleared
To: Ming Lei <ming.lei@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>, Jens Axboe <axboe@kernel.dk>,
 Pavel Begunkov <asml.silence@gmail.com>, linux-kernel@vger.kernel.org,
 linux-block@vger.kernel.org
References: <20240606135723.2794-1-yang.yang@vivo.com>
 <ZmLWfJ7OFgj+yoxV@fedora>
Content-Language: en-US
In-Reply-To: <ZmLWfJ7OFgj+yoxV@fedora>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
MIME-Version: 1.0
X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1
X-MS-Exchange-AntiSpam-MessageData-0:
	=?utf-8?B?anJFMkk1VHVLc1Z4bmJXUkUzN0poMklZZ2pUNUJsUURwUnZ5bk0zeXJ2RlNB?=
 =?utf-8?B?NUJtMWdoZFRpdmx3WEhRcWRCbDhXNEQxRmpYdHMwL0FiZDI4ZWpuczJUMko0?=
 =?utf-8?B?NURzTE1zUSt2SVBrZGE5MHhVcXRLZUZoODZncFdyYmwzdUR3WE5xVXl3Nkt0?=
 =?utf-8?B?S3NnME43eUI4OVprYS85WE82dE93TytrcHpHMFJQRDUyZ2ROdml3QjU1MmRV?=
 =?utf-8?B?RUlBd0RyYjBUNlRUVGxjUG9wVERkOWc5clo0Z2pGZktBaW9XZlNuaEFwODJI?=
 =?utf-8?B?bTZaOHJDcmRZNk9HRUFkQ0lMUDluTUNYOTZXcDYvQ2tsaFlIYnFiODF4Ky9p?=
 =?utf-8?B?Ny9WTzBvVWFNSWV4M1pwQmcwdE9lcFc3dWIvcXo3Q1E5VUVOVytEWUxHYzRv?=
 =?utf-8?B?RFdiV04wQ3lVNms4VlhiQnR3YzBNeXlLUXFmRUxEMThzakRYSHE3bWhBcExF?=
 =?utf-8?B?emZDWjJRREd1R2ZUazNQTzZWRndiaE1vazFGWjBZZGhYYytRekMwelFFK1dW?=
 =?utf-8?B?Y2JWdWpsRlg2RmhFNmdxdTdOeWtQeWVOYm1LbGpHdy9mVktpRUFJaVpaTHp0?=
 =?utf-8?B?ZGZZVmZtSjVnd1kzeEx0MzFuZW5Sam9zdUk0cEZBZCt5KzJ2S3RDa3VjcUVH?=
 =?utf-8?B?SldaeVlST0d1cUE1L1BKOUdpM0JlZWxhSTMzL1N4dzBod0lFaGt4WDZ1LzFC?=
 =?utf-8?B?S2NGOHlHQXJSclVydVFyUEw1TEp3am5nTFpwV2ZYbW1ROUtLNE5TVU9HTngv?=
 =?utf-8?B?azRrRzlqZnJKbHM0YjJZVkVxRm95ck8rNklqS1ZwQmo2ZXA4U3p5Sitza1lj?=
 =?utf-8?B?aFZmMGtXZlZxNituaklkMHdtV0t2Ky8xMWJQZEtwRFhzRUN6dU9CMlgyUkgv?=
 =?utf-8?B?TS94QjgxalRkMXZMMWNUSVdDMXZnckZ3dHFLd3lOQWMyZnMyMHZDUFNCZmVv?=
 =?utf-8?B?VkN5VXpGYzdhNzhyVFlZVHlJaGRweUt2ZVk0WHZQa3VLMjhyVFJLR0pOYmU3?=
 =?utf-8?B?SzhxUjJuSnplcTRRdTJYaXN5Mmx6NncxckVKVnVuT0pyaGpEUmhud3dzeG55?=
 =?utf-8?B?MEZTUzVYSXhnTmJQbHFWL3dlVTBWaVF6NGp3eG9uSDNJenREMy9mTUE0cGEv?=
 =?utf-8?B?TVFBSTdlNXgydzhmbm43dG1tMm8xcSs2cTFmc1dGTEZzRVpuL0Fyd2dENzNG?=
 =?utf-8?B?cm44QkJFaEVtR0NqZlV6SDljc1VGNm9oUWxHWTg3VjZRU2JLRHd5ZzZKakl0?=
 =?utf-8?B?QVhVMmthYkhUZmJJcFE1NGxGT2RtcXl3djYwNW1WOUc4dEwwQ2Npc3VQKzJY?=
 =?utf-8?B?a1g5TEdzMjY5eStac0t0V2RuSjRNUUJMbjJkYk9XcHNhQTNyanU4RXg2VmRV?=
 =?utf-8?B?RG1RNHhEbkZDM29IU3hzNTJkUXNtMXZuT3JkcmVZcTBJMGhRVWt5ck9PbW90?=
 =?utf-8?B?VXRFKys1bFFiR25JUHloUXNCNFUyY2s3MEt2ZVJoRmdsY3NHbXJZakRpY0NO?=
 =?utf-8?B?d3VKNXBMZktadW9zckdmdHYwbTYyUTVOZktzY1N0L2toaHZtRFJvTWEwcDB2?=
 =?utf-8?B?RURyNk5uTVpnL1FvMzFOZTltUlVtRmxPVEtyN2RHMmZXMDFWMGduZUtPL1F3?=
 =?utf-8?B?MWdaeWRrSC9kYVZBaGhKaTBhaVNaR1hiVHdZZEZrK1Z4MitvRExiV0hVRitZ?=
 =?utf-8?B?YkRpV1d4TXlNUDJ0Y1dYTjBtd0ZjRktjV0xITWFlZGNQcWhrRW80MDRWd1N1?=
 =?utf-8?B?dEh2WElkVW4zYTl4Y25oTEVuUGFianQyM3RrOHVabFd0S2NDNDVhSVhmYUdL?=
 =?utf-8?B?cGF2UURFQ1ZvMUtIZFEzUDdDdUl1R0VzenF6ZG81NG1zbldIeEllQ1ROY2ha?=
 =?utf-8?B?QXhQMTlSTDIxR01uOE5Hb3czSGt5T3NHOWlscG9OMVluazIvMHNnNC8veGNR?=
 =?utf-8?B?eE12VTdwSW9odnVVdnYwamwyOWZHL2ZhMGtyUUszbnZ5bER1V29SYVVqaGxB?=
 =?utf-8?B?MDN4MUF3SG1PQ3VNeldTQTgyYkNzYUpNTmxXSldTd1UyWTByVkVLUFF6dXBt?=
 =?utf-8?B?dU9zeVFhTjFsWmcvWjZhNUFtbUlKMWRnY3lPeGlVUzFNc2h6OW9uWU56UWZq?=
 =?utf-8?Q?6XUzyZ/C5b87EtuAjorz4orBi?=
X-OriginatorOrg: vivo.com
X-MS-Exchange-CrossTenant-Network-Message-Id: e145f361-1ea8-4fcf-1185-08dc8a11fe68
X-MS-Exchange-CrossTenant-AuthSource: KL1PR06MB7401.apcprd06.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 11 Jun 2024 12:28:26.8723
 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: 923e42dc-48d5-4cbe-b582-1a797a6412ed
X-MS-Exchange-CrossTenant-MailboxType: HOSTED
X-MS-Exchange-CrossTenant-UserPrincipalName: eDWJHQ0oHu1DWfXBccwVkK637k1/MK8A+rDPSuQG0n91k8gNBQ6sbNd8/kb7EA0bNJvBm6Gk7fIPYTRACchHtA==
X-MS-Exchange-Transport-CrossTenantHeadersStamped: SEZPR06MB5342

On 2024/6/7 17:44, Ming Lei wrote:
> On Thu, Jun 06, 2024 at 09:57:21PM +0800, Yang Yang wrote:
>> Configuration for sbq:
>>    depth=64, wake_batch=6, shift=6, map_nr=1
>>
>> 1. There are 64 requests in progress:
>>    map->word = 0xFFFFFFFFFFFFFFFF
>> 2. After all the 64 requests complete, and no more requests come:
>>    map->word = 0xFFFFFFFFFFFFFFFF, map->cleared = 0xFFFFFFFFFFFFFFFF
>> 3. Now two tasks try to allocate requests:
>>    T1:                                       T2:
>>    __blk_mq_get_tag                          .
>>    __sbitmap_queue_get                       .
>>    sbitmap_get                               .
>>    sbitmap_find_bit                          .
>>    sbitmap_find_bit_in_word                  .
>>    __sbitmap_get_word  -> nr=-1              __blk_mq_get_tag
>>    sbitmap_deferred_clear                    __sbitmap_queue_get
>>    /* map->cleared=0xFFFFFFFFFFFFFFFF */     sbitmap_find_bit
>>      if (!READ_ONCE(map->cleared))           sbitmap_find_bit_in_word
>>        return false;                         __sbitmap_get_word -> nr=-1
>>      mask = xchg(&map->cleared, 0)           sbitmap_deferred_clear
>>      atomic_long_andnot()                    /* map->cleared=0 */
>>                                                if (!(map->cleared))
>>                                                  return false;
>>                                       /*
>>                                        * map->cleared is cleared by T1
>>                                        * T2 fail to acquire the tag
>>                                        */
>>
>> 4. T2 is the sole tag waiter. When T1 puts the tag, T2 cannot be woken
>> up due to the wake_batch being set at 6. If no more requests come, T1
>> will wait here indefinitely.
>>
>> This patch achieves two purposes:
>> First:
>> Check on ->cleared and update on both ->cleared and ->word need to be
>> done atomically, and using spinlock could be the simplest solution.
>> So revert commit 661d4f55a794 ("sbitmap: remove swap_lock"), which
>> may cause potential race.
>>
>> Second:
>> Add extra check in sbitmap_deferred_clear(), to identify whether
>> map->cleared is cleared by another task after failing to get a tag.
>>
>> Fixes: 661d4f55a794 ("sbitmap: remove swap_lock")
>> Signed-off-by: Yang Yang <yang.yang@vivo.com>
>>
>> ---
>> Changes from v2:
>>    - Modify commit message by suggestion
>>    - Add extra check in sbitmap_deferred_clear() by suggestion
>> Changes from v1:
>>    - simply revert commit 661d4f55a794 ("sbitmap: remove swap_lock")
>> ---
>>   include/linux/sbitmap.h |  5 +++++
>>   lib/sbitmap.c           | 28 +++++++++++++++++++++-------
>>   2 files changed, 26 insertions(+), 7 deletions(-)
>>
>> diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h
>> index d662cf136021..ec0b0e73c906 100644
>> --- a/include/linux/sbitmap.h
>> +++ b/include/linux/sbitmap.h
>> @@ -36,6 +36,11 @@ struct sbitmap_word {
>>   	 * @cleared: word holding cleared bits
>>   	 */
>>   	unsigned long cleared ____cacheline_aligned_in_smp;
>> +
>> +	/**
>> +	 * @swap_lock: Held while swapping word <-> cleared
>> +	 */
>> +	spinlock_t swap_lock;
>>   } ____cacheline_aligned_in_smp;
>>   
>>   /**
>> diff --git a/lib/sbitmap.c b/lib/sbitmap.c
>> index 1e453f825c05..06b837311e03 100644
>> --- a/lib/sbitmap.c
>> +++ b/lib/sbitmap.c
>> @@ -60,12 +60,19 @@ static inline void update_alloc_hint_after_get(struct sbitmap *sb,
>>   /*
>>    * See if we have deferred clears that we can batch move
>>    */
>> -static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
>> +static inline bool sbitmap_deferred_clear(struct sbitmap_word *map,
>> +		unsigned int depth)
>>   {
>>   	unsigned long mask;
>> +	unsigned long flags;
>> +	bool ret;
>>   
>> -	if (!READ_ONCE(map->cleared))
>> -		return false;
>> +	spin_lock_irqsave(&map->swap_lock, flags);
>> +
>> +	if (!map->cleared) {
>> +		ret = find_first_zero_bit(&map->word, depth) >= depth ? false : true;
>> +		goto out_unlock;
>> +	}
> 
> Direct check over map->word should be more efficient than find_first_zero_bit():
> 
>    		if (READ_ONCE(map->word) == (1UL << depth) - 1)
> 			ret = false;
> 		else
> 			ret = true;

Got it. A little modified as follows, because:
1. The depth may be shallow_depth, which is less than __map_depth()
2. (1UL << depth) may overflow

  72 #if BITS_PER_LONG == 64
  73         unsigned long word_mask = U64_MAX >> (BITS_PER_LONG - depth);
  74 #else
  75         unsigned long word_mask = U32_MAX >> (BITS_PER_LONG - depth);
  76 #endif
  77
  78         if ((READ_ONCE(map->word) & word_mask) == word_mask)
  79             ret = false;
  80         else
  81             ret = true;

>>   
>>   	/*
>>   	 * First get a stable cleared mask, setting the old mask to 0.
>> @@ -77,7 +84,10 @@ static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
>>   	 */
>>   	atomic_long_andnot(mask, (atomic_long_t *)&map->word);
>>   	BUILD_BUG_ON(sizeof(atomic_long_t) != sizeof(map->word));
>> -	return true;
>> +	ret = true;
>> +out_unlock:
>> +	spin_unlock_irqrestore(&map->swap_lock, flags);
>> +	return ret;
>>   }
>>   
>>   int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
>> @@ -85,6 +95,7 @@ int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
>>   		      bool alloc_hint)
>>   {
>>   	unsigned int bits_per_word;
>> +	int i;
>>   
>>   	if (shift < 0)
>>   		shift = sbitmap_calculate_shift(depth);
>> @@ -116,6 +127,9 @@ int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
>>   		return -ENOMEM;
>>   	}
>>   
>> +	for (i = 0; i < sb->map_nr; i++)
>> +		spin_lock_init(&sb->map[i].swap_lock);
>> +
>>   	return 0;
>>   }
>>   EXPORT_SYMBOL_GPL(sbitmap_init_node);
>> @@ -126,7 +140,7 @@ void sbitmap_resize(struct sbitmap *sb, unsigned int depth)
>>   	unsigned int i;
>>   
>>   	for (i = 0; i < sb->map_nr; i++)
>> -		sbitmap_deferred_clear(&sb->map[i]);
>> +		sbitmap_deferred_clear(&sb->map[i], depth);
> 
> The above 'depth' is whole sbitmap depth, and you need to figure out
> word depth from __map_depth().

Got it.

Thanks.