[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3813654.1674473320@warthog.procyon.org.uk>
Date: Mon, 23 Jan 2023 11:28:40 +0000
From: David Howells <dhowells@...hat.com>
To: Christoph Hellwig <hch@...radead.org>
Cc: dhowells@...hat.com, Al Viro <viro@...iv.linux.org.uk>,
Matthew Wilcox <willy@...radead.org>,
Jens Axboe <axboe@...nel.dk>, Jan Kara <jack@...e.cz>,
Jeff Layton <jlayton@...nel.org>,
Logan Gunthorpe <logang@...tatee.com>,
linux-fsdevel@...r.kernel.org, linux-block@...r.kernel.org,
linux-kernel@...r.kernel.org, Christoph Hellwig <hch@....de>
Subject: Re: [PATCH v7 6/8] block: Make bio structs pin pages rather than ref'ing if appropriate
In this function:
void __bio_release_pages(struct bio *bio, bool mark_dirty)
{
unsigned int gup_flags = bio_to_gup_flags(bio);
struct bvec_iter_all iter_all;
struct bio_vec *bvec;
bio_for_each_segment_all(bvec, bio, iter_all) {
if (mark_dirty && !PageCompound(bvec->bv_page))
set_page_dirty_lock(bvec->bv_page);
>>>> page_put_unpin(bvec->bv_page, gup_flags);
}
}
that ought to be a call to bio_release_page(), but the optimiser doesn't want
to inline it:-/
I found the only way I can get the compiler to properly inline it without it
repeating the calculations is to renumber the FOLL_* constants down and then
make bio_release_page() something like:
static inline __attribute__((always_inline))
void bio_release_page(struct bio *bio, struct page *page)
{
page_put_unpin(page,
((bio->bi_flags & (1 << BIO_PAGE_REFFED)) ? FOLL_GET : 0) |
((bio->bi_flags & (1 << BIO_PAGE_PINNED)) ? FOLL_PIN : 0));
}
I guess the compiler optimiser isn't perfect yet:-)
David
Powered by blists - more mailing lists