linux-ext4.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Artem Blagodarenko <artem.blagodarenko@gmail.com>
To: Alex Zhuravlev <azhuravlev@whamcloud.com>
Cc: "Theodore Y. Ts'o" <tytso@mit.edu>,
	"linux-ext4@vger.kernel.org" <linux-ext4@vger.kernel.org>
Subject: Re: [RFC] improve malloc for large filesystems
Date: Thu, 21 Nov 2019 11:30:38 +0300	[thread overview]
Message-ID: <4EB2303A-01A3-49AC-B713-195126DB621B@gmail.com> (raw)
In-Reply-To: <9E04C147-D878-45CE-8473-EF8C67FE4E86@whamcloud.com>

Hello Alex,

Code looks good, but I have objections about the approach.

512TB disk with 4k block size have 4194304 groups. So 4k groups is only ~0.01% of whole disk.
Can we make decision to break search and get minimum blocks based on such limited data.
I am not sure that spending some time to find good group is worse then allocate blocks without 
optimisation. Especially, if disk is quite free and there are a lot of free block groups.

Best regards,
Artem Blagodarenko.
> On 21 Nov 2019, at 10:03, Alex Zhuravlev <azhuravlev@whamcloud.com> wrote:
> 
> 
> 
>> On 20 Nov 2019, at 21:13, Theodore Y. Ts'o <tytso@mit.edu> wrote:
>> 
>> Hi Alex,
>> 
>> A couple of comments.  First, please separate this patch so that these
>> two separate pieces of functionality can be reviewed and tested
>> separately:
>> 
> 
> This is the first patch of the series.
> 
> Thanks, Alex
> 
> From 81c4b3b5a17d94525bbc6d2d89b20f6618b05bc6 Mon Sep 17 00:00:00 2001
> From: Alex Zhuravlev <bzzz@whamcloud.com>
> Date: Thu, 21 Nov 2019 09:53:13 +0300
> Subject: [PATCH 1/2] ext4: limit scanning for a good group
> 
> at first two rounds to prevent situation when 10x-100x thousand
> of groups are scanned, especially non-initialized groups.
> 
> Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
> ---
> fs/ext4/ext4.h    |  2 ++
> fs/ext4/mballoc.c | 14 ++++++++++++--
> fs/ext4/sysfs.c   |  4 ++++
> 3 files changed, 18 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 03db3e71676c..d4e47fdad87c 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -1480,6 +1480,8 @@ struct ext4_sb_info {
> 	/* where last allocation was done - for stream allocation */
> 	unsigned long s_mb_last_group;
> 	unsigned long s_mb_last_start;
> +	unsigned int s_mb_toscan0;
> +	unsigned int s_mb_toscan1;
> 
> 	/* stats for buddy allocator */
> 	atomic_t s_bal_reqs;	/* number of reqs with len > 1 */
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index a3e2767bdf2f..cebd7d8df0b8 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -2098,7 +2098,7 @@ static int ext4_mb_good_group(struct ext4_allocation_context *ac,
> static noinline_for_stack int
> ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
> {
> -	ext4_group_t ngroups, group, i;
> +	ext4_group_t ngroups, toscan, group, i;
> 	int cr;
> 	int err = 0, first_err = 0;
> 	struct ext4_sb_info *sbi;
> @@ -2169,7 +2169,15 @@ ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
> 		 */
> 		group = ac->ac_g_ex.fe_group;
> 
> -		for (i = 0; i < ngroups; group++, i++) {
> +		/* limit number of groups to scan at the first two rounds
> +		 * when we hope to find something really good */
> +		toscan = ngroups;
> +		if (cr == 0)
> +			toscan = sbi->s_mb_toscan0;
> +		else if (cr == 1)
> +			toscan = sbi->s_mb_toscan1;
> +
> +		for (i = 0; i < toscan; group++, i++) {
> 			int ret = 0;
> 			cond_resched();
> 			/*
> @@ -2872,6 +2880,8 @@ void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid)
> 			bio_put(discard_bio);
> 		}
> 	}
> +	sbi->s_mb_toscan0 = 1024;
> +	sbi->s_mb_toscan1 = 4096;
> 
> 	list_for_each_entry_safe(entry, tmp, &freed_data_list, efd_list)
> 		ext4_free_data_in_buddy(sb, entry);
> diff --git a/fs/ext4/sysfs.c b/fs/ext4/sysfs.c
> index eb1efad0e20a..c96ee20f5487 100644
> --- a/fs/ext4/sysfs.c
> +++ b/fs/ext4/sysfs.c
> @@ -198,6 +198,8 @@ EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
> EXT4_ATTR(first_error_time, 0444, first_error_time);
> EXT4_ATTR(last_error_time, 0444, last_error_time);
> EXT4_ATTR(journal_task, 0444, journal_task);
> +EXT4_RW_ATTR_SBI_UI(mb_toscan0, s_mb_toscan0);
> +EXT4_RW_ATTR_SBI_UI(mb_toscan1, s_mb_toscan1);
> 
> static unsigned int old_bump_val = 128;
> EXT4_ATTR_PTR(max_writeback_mb_bump, 0444, pointer_ui, &old_bump_val);
> @@ -228,6 +230,8 @@ static struct attribute *ext4_attrs[] = {
> 	ATTR_LIST(first_error_time),
> 	ATTR_LIST(last_error_time),
> 	ATTR_LIST(journal_task),
> +	ATTR_LIST(mb_toscan0),
> +	ATTR_LIST(mb_toscan1),
> 	NULL,
> };
> ATTRIBUTE_GROUPS(ext4);
> -- 
> 2.20.1
> 
> 


  reply	other threads:[~2019-11-21  8:30 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-11-20 10:35 [RFC] improve malloc for large filesystems Alex Zhuravlev
2019-11-20 11:56 ` Artem Blagodarenko
2019-11-20 18:33   ` Alex Zhuravlev
2019-11-20 18:13 ` Theodore Y. Ts'o
2019-11-20 18:22   ` Alex Zhuravlev
2019-11-21  7:03   ` Alex Zhuravlev
2019-11-21  8:30     ` Artem Blagodarenko [this message]
2019-11-21  8:52       ` Alex Zhuravlev
2019-11-21  9:18         ` Artem Blagodarenko
2019-11-21 14:41           ` Alex Zhuravlev
2019-11-25 21:39             ` Andreas Dilger
2019-12-02  8:46               ` Alex Zhuravlev
2019-11-21  7:03   ` Alex Zhuravlev

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=4EB2303A-01A3-49AC-B713-195126DB621B@gmail.com \
    --to=artem.blagodarenko@gmail.com \
    --cc=azhuravlev@whamcloud.com \
    --cc=linux-ext4@vger.kernel.org \
    --cc=tytso@mit.edu \
    /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).