linux-block.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Christoph Hellwig <hch@infradead.org>
To: Jens Axboe <axboe@kernel.dk>
Cc: io-uring@vger.kernel.org, linux-block@vger.kernel.org, hch@infradead.org
Subject: Re: [PATCH 3/6] bio: add allocation cache abstraction
Date: Thu, 12 Aug 2021 09:01:35 +0200	[thread overview]
Message-ID: <YRTHTyz+tlRuGv2i@infradead.org> (raw)
In-Reply-To: <20210811193533.766613-4-axboe@kernel.dk>

On Wed, Aug 11, 2021 at 01:35:30PM -0600, Jens Axboe wrote:
> +	struct bio *bio;
> +	unsigned int i;
> +
> +	i = 0;

Initialize at declaration time?

> +static inline bool __bio_put(struct bio *bio)
> +{
> +	if (!bio_flagged(bio, BIO_REFFED))
> +		return true;
> +
> +	BIO_BUG_ON(!atomic_read(&bio->__bi_cnt));
> +
> +	/*
> +	 * last put frees it
> +	 */
> +	return atomic_dec_and_test(&bio->__bi_cnt);
> +}

Please avoid this helper, we can trivially do the check inside of bio_put:

	if (bio_flagged(bio, BIO_REFFED)) {
		BIO_BUG_ON(!atomic_read(&bio->__bi_cnt));
		if (!atomic_dec_and_test(&bio->__bi_cnt))
			return;
	}

> -			bio_free(bio);
> +	if (bio_flagged(bio, BIO_PERCPU_CACHE)) {
> +		struct bio_alloc_cache *cache;
> +
> +		bio_uninit(bio);
> +		cache = per_cpu_ptr(bio->bi_pool->cache, get_cpu());
> +		bio_list_add_head(&cache->free_list, bio);
> +		cache->nr++;
> +		if (cache->nr > ALLOC_CACHE_MAX + ALLOC_CACHE_SLACK)

Folding the increment as a prefix here would make the increment and test
semantics a little more obvious.

> +struct bio *bio_alloc_kiocb(struct kiocb *kiocb, gfp_t gfp,
> +			    unsigned short nr_vecs, struct bio_set *bs)
> +{
> +	struct bio_alloc_cache *cache = NULL;
> +	struct bio *bio;
> +
> +	if (!(kiocb->ki_flags & IOCB_ALLOC_CACHE) || nr_vecs > BIO_INLINE_VECS)
> +		goto normal_alloc;
> +
> +	cache = per_cpu_ptr(bs->cache, get_cpu());
> +	bio = bio_list_pop(&cache->free_list);
> +	if (bio) {
> +		cache->nr--;
> +		put_cpu();
> +		bio_init(bio, nr_vecs ? bio->bi_inline_vecs : NULL, nr_vecs);
> +		bio_set_flag(bio, BIO_PERCPU_CACHE);
> +		return bio;
> +	}
> +	put_cpu();
> +normal_alloc:
> +	bio = bio_alloc_bioset(gfp, nr_vecs, bs);
> +	if (cache)
> +		bio_set_flag(bio, BIO_PERCPU_CACHE);
> +	return bio;

The goto here is pretty obsfucating and adds an extra patch to the fast
path.

Also I don't think we need the gfp argument here at all - it should
always be GFP_KERNEL.  In fact I plan to kill these arguments from much
of the block layer as the places that need NOIO of NOFS semantics can
and should move to set that in the per-thread context.

> -static inline struct bio *bio_list_pop(struct bio_list *bl)
> +static inline void bio_list_del_head(struct bio_list *bl, struct bio *head)
>  {
> -	struct bio *bio = bl->head;
> -
> -	if (bio) {
> +	if (head) {
>  		bl->head = bl->head->bi_next;
>  		if (!bl->head)
>  			bl->tail = NULL;
>  
> -		bio->bi_next = NULL;
> +		head->bi_next = NULL;
>  	}
> +}
> +
> +static inline struct bio *bio_list_pop(struct bio_list *bl)
> +{
> +	struct bio *bio = bl->head;
>  
> +	bio_list_del_head(bl, bio);
>  	return bio;
>  }

No need for this change.

>
>  
> @@ -699,6 +706,12 @@ struct bio_set {
>  	struct kmem_cache *bio_slab;
>  	unsigned int front_pad;
>  
> +	/*
> +	 * per-cpu bio alloc cache and notifier
> +	 */
> +	struct bio_alloc_cache __percpu *cache;
> +	struct hlist_node cpuhp_dead;
> +

I'd keep the hotplug list entry at the end instead of bloating the
cache line used in the fast path.

  reply	other threads:[~2021-08-12  7:02 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-11 19:35 [PATCHSET v4 0/6] Enable bio recycling for polled IO Jens Axboe
2021-08-11 19:35 ` [PATCH 1/6] bio: optimize initialization of a bio Jens Axboe
2021-08-12  6:51   ` Christoph Hellwig
2021-08-12 15:29     ` Jens Axboe
2021-08-11 19:35 ` [PATCH 2/6] fs: add kiocb alloc cache flag Jens Axboe
2021-08-12  6:54   ` Christoph Hellwig
2021-08-12 14:52     ` Jens Axboe
2021-08-11 19:35 ` [PATCH 3/6] bio: add allocation cache abstraction Jens Axboe
2021-08-12  7:01   ` Christoph Hellwig [this message]
2021-08-12 15:08     ` Jens Axboe
2021-08-12 15:18       ` Christoph Hellwig
2021-08-12 15:26         ` Jens Axboe
2021-08-11 19:35 ` [PATCH 4/6] block: clear BIO_PERCPU_CACHE flag if polling isn't supported Jens Axboe
2021-08-11 19:35 ` [PATCH 5/6] io_uring: enable use of bio alloc cache Jens Axboe
2021-08-11 19:35 ` [PATCH 6/6] block: enable use of bio allocation cache Jens Axboe
2021-08-12  7:04   ` Christoph Hellwig
2021-08-12 14:52     ` Jens Axboe
2021-08-12 15:41 [PATCHSET v5 0/6] Enable bio recycling for polled IO Jens Axboe
2021-08-12 15:41 ` [PATCH 3/6] bio: add allocation cache abstraction Jens Axboe
2021-08-12 16:32   ` Christoph Hellwig
2021-08-12 16:39     ` Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=YRTHTyz+tlRuGv2i@infradead.org \
    --to=hch@infradead.org \
    --cc=axboe@kernel.dk \
    --cc=io-uring@vger.kernel.org \
    --cc=linux-block@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).