Date: Fri, 20 Aug 2021 21:18:22 GMT From: Maxim Sobolev <sobomax@FreeBSD.org> To: src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org Subject: git: 0d13f5343faf - main - Only trigger read-ahead if two adjacent blocks have been requested. Message-ID: <202108202118.17KLIMlx055748@gitrepo.freebsd.org>
next in thread | raw e-mail | index | archive | help
The branch main has been updated by sobomax: URL: https://cgit.FreeBSD.org/src/commit/?id=0d13f5343fafbf3067ffc33a507ffca0375c4417 commit 0d13f5343fafbf3067ffc33a507ffca0375c4417 Author: Maxim Sobolev <sobomax@FreeBSD.org> AuthorDate: 2021-08-20 19:33:51 +0000 Commit: Maxim Sobolev <sobomax@FreeBSD.org> CommitDate: 2021-08-20 21:08:01 +0000 Only trigger read-ahead if two adjacent blocks have been requested. The change makes block caching algorithm to work better for remote media on low-BW/high-delay links. This cuts boot time over IP KVMs noticeably, since the initialization stage reads bunch of small 4th (and now lua) files that are not in the same cache stripe (usually), thus wasting lot of bandwidth and increasing latency even further. The original regression came in 2017 with revision 87ed2b7f5. We've seen increase of time it takes for the loader to get to the kernel loading from under a minute to 10-15 minutes in many cases. Reviewed by: tsoome MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D31623 --- stand/common/bcache.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-) diff --git a/stand/common/bcache.c b/stand/common/bcache.c index a020f3c3c53c..526f9fe3fa5c 100644 --- a/stand/common/bcache.c +++ b/stand/common/bcache.c @@ -66,6 +66,7 @@ struct bcache { caddr_t bcache_data; size_t bcache_nblks; size_t ra; + daddr_t bcache_nextblkno; }; static u_int bcache_total_nblks; /* set by bcache_init */ @@ -163,6 +164,7 @@ bcache_allocate(void) } bcache_units++; bc->ra = BCACHE_READAHEAD; /* optimistic read ahead */ + bc->bcache_nextblkno = -1; return (bc); } @@ -291,6 +293,14 @@ read_strategy(void *devdata, int rw, daddr_t blk, size_t size, else ra = bc->bcache_nblks - BHASH(bc, p_blk + p_size); + /* + * Only trigger read-ahead if we detect two blocks being read + * sequentially. + */ + if ((bc->bcache_nextblkno != blk) && ra != 0) { + ra = 0; + } + if (ra != 0 && ra != bc->bcache_nblks) { /* do we have RA space? */ ra = MIN(bc->ra, ra - 1); ra = rounddown(ra, 16); /* multiple of 16 blocks */ @@ -342,8 +352,11 @@ read_strategy(void *devdata, int rw, daddr_t blk, size_t size, } done: - if ((result == 0) && (rsize != NULL)) - *rsize = size; + if (result == 0) { + if (rsize != NULL) + *rsize = size; + bc->bcache_nextblkno = blk + (size / DEV_BSIZE); + } return(result); }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202108202118.17KLIMlx055748>