2023-07-26 07:49:09

by Muchun Song

[permalink] [raw]
Subject: Re: [PATCH v2 28/47] bcache: dynamically allocate the md-bcache shrinker



On 2023/7/24 17:43, Qi Zheng wrote:
> In preparation for implementing lockless slab shrink, use new APIs to
> dynamically allocate the md-bcache shrinker, so that it can be freed
> asynchronously using kfree_rcu(). Then it doesn't need to wait for RCU
> read-side critical section when releasing the struct cache_set.
>
> Signed-off-by: Qi Zheng <[email protected]>
> ---
> drivers/md/bcache/bcache.h | 2 +-
> drivers/md/bcache/btree.c | 27 ++++++++++++++++-----------
> drivers/md/bcache/sysfs.c | 3 ++-
> 3 files changed, 19 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h
> index 5a79bb3c272f..c622bc50f81b 100644
> --- a/drivers/md/bcache/bcache.h
> +++ b/drivers/md/bcache/bcache.h
> @@ -541,7 +541,7 @@ struct cache_set {
> struct bio_set bio_split;
>
> /* For the btree cache */
> - struct shrinker shrink;
> + struct shrinker *shrink;
>
> /* For the btree cache and anything allocation related */
> struct mutex bucket_lock;
> diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c
> index fd121a61f17c..c176c7fc77d9 100644
> --- a/drivers/md/bcache/btree.c
> +++ b/drivers/md/bcache/btree.c
> @@ -667,7 +667,7 @@ static int mca_reap(struct btree *b, unsigned int min_order, bool flush)
> static unsigned long bch_mca_scan(struct shrinker *shrink,
> struct shrink_control *sc)
> {
> - struct cache_set *c = container_of(shrink, struct cache_set, shrink);
> + struct cache_set *c = shrink->private_data;
> struct btree *b, *t;
> unsigned long i, nr = sc->nr_to_scan;
> unsigned long freed = 0;
> @@ -734,7 +734,7 @@ static unsigned long bch_mca_scan(struct shrinker *shrink,
> static unsigned long bch_mca_count(struct shrinker *shrink,
> struct shrink_control *sc)
> {
> - struct cache_set *c = container_of(shrink, struct cache_set, shrink);
> + struct cache_set *c = shrink->private_data;
>
> if (c->shrinker_disabled)
> return 0;
> @@ -752,8 +752,8 @@ void bch_btree_cache_free(struct cache_set *c)
>
> closure_init_stack(&cl);
>
> - if (c->shrink.list.next)
> - unregister_shrinker(&c->shrink);
> + if (c->shrink)
> + shrinker_unregister(c->shrink);
>
> mutex_lock(&c->bucket_lock);
>
> @@ -828,14 +828,19 @@ int bch_btree_cache_alloc(struct cache_set *c)
> c->verify_data = NULL;
> #endif
>
> - c->shrink.count_objects = bch_mca_count;
> - c->shrink.scan_objects = bch_mca_scan;
> - c->shrink.seeks = 4;
> - c->shrink.batch = c->btree_pages * 2;
> + c->shrink = shrinker_alloc(0, "md-bcache:%pU", c->set_uuid);
> + if (!c->shrink) {
> + pr_warn("bcache: %s: could not allocate shrinker\n", __func__);
> + return -ENOMEM;

Seems you have cheanged the semantic of this. In the past,
it is better to have a shrinker, but now it becomes a mandatory.
Right? I don't know if it is acceptable. From my point of view,
just do the cleanup, don't change any behaviour.

> + }
> +
> + c->shrink->count_objects = bch_mca_count;
> + c->shrink->scan_objects = bch_mca_scan;
> + c->shrink->seeks = 4;
> + c->shrink->batch = c->btree_pages * 2;
> + c->shrink->private_data = c;
>
> - if (register_shrinker(&c->shrink, "md-bcache:%pU", c->set_uuid))
> - pr_warn("bcache: %s: could not register shrinker\n",
> - __func__);
> + shrinker_register(c->shrink);
>
> return 0;
> }
> diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c
> index 0e2c1880f60b..45d8af755de6 100644
> --- a/drivers/md/bcache/sysfs.c
> +++ b/drivers/md/bcache/sysfs.c
> @@ -866,7 +866,8 @@ STORE(__bch_cache_set)
>
> sc.gfp_mask = GFP_KERNEL;
> sc.nr_to_scan = strtoul_or_return(buf);
> - c->shrink.scan_objects(&c->shrink, &sc);
> + if (c->shrink)
> + c->shrink->scan_objects(c->shrink, &sc);
> }
>
> sysfs_strtoul_clamp(congested_read_threshold_us,



2023-07-26 09:51:12

by Qi Zheng

[permalink] [raw]
Subject: Re: [PATCH v2 28/47] bcache: dynamically allocate the md-bcache shrinker



On 2023/7/26 15:32, Muchun Song wrote:
>
>
> On 2023/7/24 17:43, Qi Zheng wrote:
>> In preparation for implementing lockless slab shrink, use new APIs to
>> dynamically allocate the md-bcache shrinker, so that it can be freed
>> asynchronously using kfree_rcu(). Then it doesn't need to wait for RCU
>> read-side critical section when releasing the struct cache_set.
>>
>> Signed-off-by: Qi Zheng <[email protected]>
>> ---
>>   drivers/md/bcache/bcache.h |  2 +-
>>   drivers/md/bcache/btree.c  | 27 ++++++++++++++++-----------
>>   drivers/md/bcache/sysfs.c  |  3 ++-
>>   3 files changed, 19 insertions(+), 13 deletions(-)
>>
>> diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h
>> index 5a79bb3c272f..c622bc50f81b 100644
>> --- a/drivers/md/bcache/bcache.h
>> +++ b/drivers/md/bcache/bcache.h
>> @@ -541,7 +541,7 @@ struct cache_set {
>>       struct bio_set        bio_split;
>>       /* For the btree cache */
>> -    struct shrinker        shrink;
>> +    struct shrinker        *shrink;
>>       /* For the btree cache and anything allocation related */
>>       struct mutex        bucket_lock;
>> diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c
>> index fd121a61f17c..c176c7fc77d9 100644
>> --- a/drivers/md/bcache/btree.c
>> +++ b/drivers/md/bcache/btree.c
>> @@ -667,7 +667,7 @@ static int mca_reap(struct btree *b, unsigned int
>> min_order, bool flush)
>>   static unsigned long bch_mca_scan(struct shrinker *shrink,
>>                     struct shrink_control *sc)
>>   {
>> -    struct cache_set *c = container_of(shrink, struct cache_set,
>> shrink);
>> +    struct cache_set *c = shrink->private_data;
>>       struct btree *b, *t;
>>       unsigned long i, nr = sc->nr_to_scan;
>>       unsigned long freed = 0;
>> @@ -734,7 +734,7 @@ static unsigned long bch_mca_scan(struct shrinker
>> *shrink,
>>   static unsigned long bch_mca_count(struct shrinker *shrink,
>>                      struct shrink_control *sc)
>>   {
>> -    struct cache_set *c = container_of(shrink, struct cache_set,
>> shrink);
>> +    struct cache_set *c = shrink->private_data;
>>       if (c->shrinker_disabled)
>>           return 0;
>> @@ -752,8 +752,8 @@ void bch_btree_cache_free(struct cache_set *c)
>>       closure_init_stack(&cl);
>> -    if (c->shrink.list.next)
>> -        unregister_shrinker(&c->shrink);
>> +    if (c->shrink)
>> +        shrinker_unregister(c->shrink);
>>       mutex_lock(&c->bucket_lock);
>> @@ -828,14 +828,19 @@ int bch_btree_cache_alloc(struct cache_set *c)
>>           c->verify_data = NULL;
>>   #endif
>> -    c->shrink.count_objects = bch_mca_count;
>> -    c->shrink.scan_objects = bch_mca_scan;
>> -    c->shrink.seeks = 4;
>> -    c->shrink.batch = c->btree_pages * 2;
>> +    c->shrink = shrinker_alloc(0, "md-bcache:%pU", c->set_uuid);
>> +    if (!c->shrink) {
>> +        pr_warn("bcache: %s: could not allocate shrinker\n", __func__);
>> +        return -ENOMEM;
>
> Seems you have cheanged the semantic of this. In the past,
> it is better to have a shrinker, but now it becomes a mandatory.
> Right? I don't know if it is acceptable. From my point of view,
> just do the cleanup, don't change any behaviour.

Oh, should return 0 here, will do.

>
>> +    }
>> +
>> +    c->shrink->count_objects = bch_mca_count;
>> +    c->shrink->scan_objects = bch_mca_scan;
>> +    c->shrink->seeks = 4;
>> +    c->shrink->batch = c->btree_pages * 2;
>> +    c->shrink->private_data = c;
>> -    if (register_shrinker(&c->shrink, "md-bcache:%pU", c->set_uuid))
>> -        pr_warn("bcache: %s: could not register shrinker\n",
>> -                __func__);
>> +    shrinker_register(c->shrink);
>>       return 0;
>>   }
>> diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c
>> index 0e2c1880f60b..45d8af755de6 100644
>> --- a/drivers/md/bcache/sysfs.c
>> +++ b/drivers/md/bcache/sysfs.c
>> @@ -866,7 +866,8 @@ STORE(__bch_cache_set)
>>           sc.gfp_mask = GFP_KERNEL;
>>           sc.nr_to_scan = strtoul_or_return(buf);
>> -        c->shrink.scan_objects(&c->shrink, &sc);
>> +        if (c->shrink)
>> +            c->shrink->scan_objects(c->shrink, &sc);
>>       }
>>       sysfs_strtoul_clamp(congested_read_threshold_us,
>