Message ID | 1378973304-11693-10-git-send-email-quwenruo@cn.fujitsu.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Thu, Sep 12, 2013 at 04:08:24PM +0800, Qu Wenruo wrote: > The original btrfs_workers uses the fs_info->thread_pool_size as the > max_active, and the previous patches followed this way. > > But the kernel workqueue has the default value(0) for workqueue, > and workqueue itself has some threshold mechanism to prevent creating > too many threads, so we should use the default value. > > Since the thread_pool_size algorithm is not used, related codes should > also be changed. Ohh, I should have seen this mail first before commenting 'max_active'. I think that some tuning work should be done on this part, according to my tests, setting max_active=0 will create ~258 worker helpers (kworker/uX:X if you set WQ_UNBOUND), this may cause too many context switches which will have an impact on performance in some cases. -liubo > > Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> > --- > fs/btrfs/disk-io.c | 12 +++++++----- > fs/btrfs/super.c | 3 +-- > 2 files changed, 8 insertions(+), 7 deletions(-) > > diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c > index a61e1fe..0446d27 100644 > --- a/fs/btrfs/disk-io.c > +++ b/fs/btrfs/disk-io.c > @@ -750,9 +750,11 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, > > unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) > { > - unsigned long limit = min_t(unsigned long, > - info->thread_pool_size, > - info->fs_devices->open_devices); > + unsigned long limit; > + limit = info->thread_pool_size ? > + min_t(unsigned long, info->thread_pool_size, > + info->fs_devices->open_devices) : > + info->fs_devices->open_devices; > return 256 * limit; > } > > @@ -2191,8 +2193,8 @@ int open_ctree(struct super_block *sb, > INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); > spin_lock_init(&fs_info->reada_lock); > > - fs_info->thread_pool_size = min_t(unsigned long, > - num_online_cpus() + 2, 8); > + /* use the default value of kernel workqueue */ > + fs_info->thread_pool_size = 0; > > INIT_LIST_HEAD(&fs_info->ordered_roots); > spin_lock_init(&fs_info->ordered_root_lock); > diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c > index 63e653c..ccf412f 100644 > --- a/fs/btrfs/super.c > +++ b/fs/btrfs/super.c > @@ -898,8 +898,7 @@ static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) > if (info->alloc_start != 0) > seq_printf(seq, ",alloc_start=%llu", > (unsigned long long)info->alloc_start); > - if (info->thread_pool_size != min_t(unsigned long, > - num_online_cpus() + 2, 8)) > + if (info->thread_pool_size) > seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); > if (btrfs_test_opt(root, COMPRESS)) { > if (info->compress_type == BTRFS_COMPRESS_ZLIB) > -- > 1.8.4 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
? 2013?09?13? 09:47, Liu Bo ??: > On Thu, Sep 12, 2013 at 04:08:24PM +0800, Qu Wenruo wrote: >> The original btrfs_workers uses the fs_info->thread_pool_size as the >> max_active, and the previous patches followed this way. >> >> But the kernel workqueue has the default value(0) for workqueue, >> and workqueue itself has some threshold mechanism to prevent creating >> too many threads, so we should use the default value. >> >> Since the thread_pool_size algorithm is not used, related codes should >> also be changed. > Ohh, I should have seen this mail first before commenting 'max_active'. > > I think that some tuning work should be done on this part, according to > my tests, setting max_active=0 will create ~258 worker helpers > (kworker/uX:X if you set WQ_UNBOUND), this may cause too many context > switches which will have an impact on performance in some cases. Yes, but the default number when using max_active=0 should be 256 (half of the WQ_DFL_ACTIVE). Also in my test(single thread), the performance and CPU usage does not change too much. So it seems that in this situation, the kernel has some control on creating kthreads. Still further max_active tunning is still quiet good. Qu > > -liubo > >> Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> >> --- >> fs/btrfs/disk-io.c | 12 +++++++----- >> fs/btrfs/super.c | 3 +-- >> 2 files changed, 8 insertions(+), 7 deletions(-) >> >> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c >> index a61e1fe..0446d27 100644 >> --- a/fs/btrfs/disk-io.c >> +++ b/fs/btrfs/disk-io.c >> @@ -750,9 +750,11 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, >> >> unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) >> { >> - unsigned long limit = min_t(unsigned long, >> - info->thread_pool_size, >> - info->fs_devices->open_devices); >> + unsigned long limit; >> + limit = info->thread_pool_size ? >> + min_t(unsigned long, info->thread_pool_size, >> + info->fs_devices->open_devices) : >> + info->fs_devices->open_devices; >> return 256 * limit; >> } >> >> @@ -2191,8 +2193,8 @@ int open_ctree(struct super_block *sb, >> INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); >> spin_lock_init(&fs_info->reada_lock); >> >> - fs_info->thread_pool_size = min_t(unsigned long, >> - num_online_cpus() + 2, 8); >> + /* use the default value of kernel workqueue */ >> + fs_info->thread_pool_size = 0; >> >> INIT_LIST_HEAD(&fs_info->ordered_roots); >> spin_lock_init(&fs_info->ordered_root_lock); >> diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c >> index 63e653c..ccf412f 100644 >> --- a/fs/btrfs/super.c >> +++ b/fs/btrfs/super.c >> @@ -898,8 +898,7 @@ static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) >> if (info->alloc_start != 0) >> seq_printf(seq, ",alloc_start=%llu", >> (unsigned long long)info->alloc_start); >> - if (info->thread_pool_size != min_t(unsigned long, >> - num_online_cpus() + 2, 8)) >> + if (info->thread_pool_size) >> seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); >> if (btrfs_test_opt(root, COMPRESS)) { >> if (info->compress_type == BTRFS_COMPRESS_ZLIB) >> -- >> 1.8.4 >> >> -- >> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in >> the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html > -- > To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >
? 2013?09?13? 11:15, Qu Wenruo ??: > ? 2013?09?13? 09:47, Liu Bo ??: >> On Thu, Sep 12, 2013 at 04:08:24PM +0800, Qu Wenruo wrote: >>> The original btrfs_workers uses the fs_info->thread_pool_size as the >>> max_active, and the previous patches followed this way. >>> >>> But the kernel workqueue has the default value(0) for workqueue, >>> and workqueue itself has some threshold mechanism to prevent creating >>> too many threads, so we should use the default value. >>> >>> Since the thread_pool_size algorithm is not used, related codes should >>> also be changed. >> Ohh, I should have seen this mail first before commenting 'max_active'. >> >> I think that some tuning work should be done on this part, according to >> my tests, setting max_active=0 will create ~258 worker helpers >> (kworker/uX:X if you set WQ_UNBOUND), this may cause too many context >> switches which will have an impact on performance in some cases. > Yes, but the default number when using max_active=0 should be 256 > (half of the WQ_DFL_ACTIVE). > > Also in my test(single thread), the performance and CPU usage does not > change too much. > So it seems that in this situation, the kernel has some control on > creating > kthreads. > > Still further max_active tunning is still quiet good. > > Qu Sorry for the last reply, according to the workqueue source code, the unbound workqueue will continually create new thread if needed. So the default value is not good for this situation. Also there are so many wq using the unbound wq, the old thread_pool_size is too small. (Maybe change some wq to bounded would be a good idea?) It would be quite nice if you have any good idea or advice about the tunning about max_active. Qu >> >> -liubo >> >>> Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> >>> --- >>> fs/btrfs/disk-io.c | 12 +++++++----- >>> fs/btrfs/super.c | 3 +-- >>> 2 files changed, 8 insertions(+), 7 deletions(-) >>> >>> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c >>> index a61e1fe..0446d27 100644 >>> --- a/fs/btrfs/disk-io.c >>> +++ b/fs/btrfs/disk-io.c >>> @@ -750,9 +750,11 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info >>> *info, struct bio *bio, >>> unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) >>> { >>> - unsigned long limit = min_t(unsigned long, >>> - info->thread_pool_size, >>> - info->fs_devices->open_devices); >>> + unsigned long limit; >>> + limit = info->thread_pool_size ? >>> + min_t(unsigned long, info->thread_pool_size, >>> + info->fs_devices->open_devices) : >>> + info->fs_devices->open_devices; >>> return 256 * limit; >>> } >>> @@ -2191,8 +2193,8 @@ int open_ctree(struct super_block *sb, >>> INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); >>> spin_lock_init(&fs_info->reada_lock); >>> - fs_info->thread_pool_size = min_t(unsigned long, >>> - num_online_cpus() + 2, 8); >>> + /* use the default value of kernel workqueue */ >>> + fs_info->thread_pool_size = 0; >>> INIT_LIST_HEAD(&fs_info->ordered_roots); >>> spin_lock_init(&fs_info->ordered_root_lock); >>> diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c >>> index 63e653c..ccf412f 100644 >>> --- a/fs/btrfs/super.c >>> +++ b/fs/btrfs/super.c >>> @@ -898,8 +898,7 @@ static int btrfs_show_options(struct seq_file >>> *seq, struct dentry *dentry) >>> if (info->alloc_start != 0) >>> seq_printf(seq, ",alloc_start=%llu", >>> (unsigned long long)info->alloc_start); >>> - if (info->thread_pool_size != min_t(unsigned long, >>> - num_online_cpus() + 2, 8)) >>> + if (info->thread_pool_size) >>> seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); >>> if (btrfs_test_opt(root, COMPRESS)) { >>> if (info->compress_type == BTRFS_COMPRESS_ZLIB) >>> -- >>> 1.8.4 >>> >>> -- >>> To unsubscribe from this list: send the line "unsubscribe >>> linux-btrfs" in >>> the body of a message to majordomo@vger.kernel.org >>> More majordomo info at http://vger.kernel.org/majordomo-info.html >> -- >> To unsubscribe from this list: send the line "unsubscribe >> linux-btrfs" in >> the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html >> > >
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index a61e1fe..0446d27 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -750,9 +750,11 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) { - unsigned long limit = min_t(unsigned long, - info->thread_pool_size, - info->fs_devices->open_devices); + unsigned long limit; + limit = info->thread_pool_size ? + min_t(unsigned long, info->thread_pool_size, + info->fs_devices->open_devices) : + info->fs_devices->open_devices; return 256 * limit; } @@ -2191,8 +2193,8 @@ int open_ctree(struct super_block *sb, INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); spin_lock_init(&fs_info->reada_lock); - fs_info->thread_pool_size = min_t(unsigned long, - num_online_cpus() + 2, 8); + /* use the default value of kernel workqueue */ + fs_info->thread_pool_size = 0; INIT_LIST_HEAD(&fs_info->ordered_roots); spin_lock_init(&fs_info->ordered_root_lock); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 63e653c..ccf412f 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -898,8 +898,7 @@ static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) if (info->alloc_start != 0) seq_printf(seq, ",alloc_start=%llu", (unsigned long long)info->alloc_start); - if (info->thread_pool_size != min_t(unsigned long, - num_online_cpus() + 2, 8)) + if (info->thread_pool_size) seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); if (btrfs_test_opt(root, COMPRESS)) { if (info->compress_type == BTRFS_COMPRESS_ZLIB)
The original btrfs_workers uses the fs_info->thread_pool_size as the max_active, and the previous patches followed this way. But the kernel workqueue has the default value(0) for workqueue, and workqueue itself has some threshold mechanism to prevent creating too many threads, so we should use the default value. Since the thread_pool_size algorithm is not used, related codes should also be changed. Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> --- fs/btrfs/disk-io.c | 12 +++++++----- fs/btrfs/super.c | 3 +-- 2 files changed, 8 insertions(+), 7 deletions(-)