GNU/Weeb Mailing List <[email protected]>
 help / color / mirror / Atom feed
From: Ammar Faizi <[email protected]>
To: Chris Mason <[email protected]>, Josef Bacik <[email protected]>,
	David Sterba <[email protected]>, Tejun Heo <[email protected]>
Cc: Ammar Faizi <[email protected]>,
	Lai Jiangshan <[email protected]>,
	Filipe Manana <[email protected]>,
	Linux Btrfs Mailing List <[email protected]>,
	Linux Kernel Mailing List <[email protected]>,
	Linux Fsdevel Mailing List <[email protected]>,
	GNU/Weeb Mailing List <[email protected]>
Subject: [RFC PATCH v1 5/6] btrfs: Adjust the default thread pool size when `wq_cpu_set` option is used
Date: Sun, 26 Feb 2023 23:02:58 +0700	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

If wq_cpu_set is specified, adjust thread_pool_size to the number of
CPUs in the set plus 2 to avoid the case where the number of CPUs in the
set is less than the default thread_pool_size + 2, which might cause the
thread pool to be starved. The thread_pool=%u mount option overrides
this adjusting behavior.

Signed-off-by: Ammar Faizi <[email protected]>
---
 fs/btrfs/super.c | 38 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)

diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 3e061ec977b014d1..34b7c5810d34d624 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -317,6 +317,32 @@ static int parse_wq_cpu_set(struct btrfs_fs_info *info, const char *mask_str)
 	return 0;
 }
 
+static void adjust_default_thread_pool_size(struct btrfs_fs_info *info)
+{
+	unsigned long old_thread_pool_size;
+	unsigned long new_thread_pool_size;
+	unsigned long total_usable_cpu = 0;
+	unsigned long cpu;
+
+	if (!btrfs_test_opt(info, WQ_CPU_SET))
+		return;
+
+	for_each_online_cpu(cpu) {
+		if (cpumask_test_cpu(cpu, info->wq_cpu_set->mask))
+			total_usable_cpu++;
+	}
+
+	old_thread_pool_size = info->thread_pool_size;
+	new_thread_pool_size = min_t(unsigned long, total_usable_cpu + 2, 8);
+
+	if (old_thread_pool_size == new_thread_pool_size)
+		return;
+
+	info->thread_pool_size = new_thread_pool_size;
+	btrfs_info(info, "adjusting thread_pool_size to %lu due to wq_cpu_set (you can override this with thread_pool=%%u option)",
+		   new_thread_pool_size);
+}
+
 /*
  * Regular mount options parser.  Everything that is needed only when
  * reading in a new superblock is parsed here.
@@ -336,6 +362,7 @@ int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 	bool saved_compress_force;
 	int no_compress = 0;
 	const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
+	bool has_thread_pool_opt = false;
 
 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 		btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
@@ -543,6 +570,7 @@ int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 				goto out;
 			}
 			info->thread_pool_size = intarg;
+			has_thread_pool_opt = true;
 			break;
 		case Opt_max_inline:
 			num = match_strdup(&args[0]);
@@ -854,6 +882,16 @@ int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 	}
 	if (!ret)
 		ret = btrfs_check_mountopts_zoned(info);
+
+	/*
+	 * If wq_cpu_set is specified, adjust thread_pool_size to the number of
+	 * CPUs in the set plus 2 to avoid the case where the number of CPUs in
+	 * the set is less than the default thread_pool_size + 2, which might
+	 * cause the thread pool to be starved. The thread_pool=%u mount option
+	 * overrides this adjusting behavior.
+	 */
+	if (!ret && !has_thread_pool_opt)
+		adjust_default_thread_pool_size(info);
 	if (!ret && !remounting) {
 		if (btrfs_test_opt(info, SPACE_CACHE))
 			btrfs_info(info, "disk space caching is enabled");
-- 
Ammar Faizi


  parent reply	other threads:[~2023-02-26 16:03 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-02-26 16:02 [RFC PATCH v1 0/6] Introducing `wq_cpu_set` mount option for btrfs Ammar Faizi
2023-02-26 16:02 ` [RFC PATCH v1 1/6] workqueue: Add set_workqueue_cpumask() helper function Ammar Faizi
2023-02-26 16:02 ` [RFC PATCH v1 2/6] btrfs: Change `mount_opt` type in `struct btrfs_fs_info` to `u64` Ammar Faizi
2023-02-26 16:02 ` [RFC PATCH v1 3/6] btrfs: Create btrfs CPU set struct and helpers Ammar Faizi
2023-02-26 16:02 ` [RFC PATCH v1 4/6] btrfs: Add wq_cpu_set=%s mount option Ammar Faizi
2023-02-26 16:02 ` Ammar Faizi [this message]
2023-02-26 16:02 ` [RFC PATCH v1 6/6] btrfs: Add `BTRFS_DEFAULT_MAX_THREAD_POOL_SIZE` macro Ammar Faizi
2023-02-26 17:01 ` [RFC PATCH v1 0/6] Introducing `wq_cpu_set` mount option for btrfs Tejun Heo
2023-02-26 18:26   ` Ammar Faizi
2023-02-26 18:29     ` Ammar Faizi
2023-02-27 10:18 ` Qu Wenruo
2023-02-27 13:42   ` Ammar Faizi
2023-02-27 23:49     ` Qu Wenruo
2023-02-27 11:02 ` Filipe Manana
     [not found]   ` <[email protected]>
2023-02-27 13:45     ` Ammar Faizi
2023-02-27 16:24       ` Roman Mamedov
2023-02-27 22:17 ` Dave Chinner
2023-02-28  8:01   ` Ammar Faizi

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 \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /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