Re: [f2fs-dev] [PATCH] f2fs: add tracepoint for f2fs iostat

From: Jaegeuk Kim
Date: Wed Apr 15 2020 - 16:56:17 EST


On 04/15, Chao Yu wrote:
> On 2020/4/14 0:16, Jaegeuk Kim wrote:
> > From: Daeho Jeong <daehojeong@xxxxxxxxxx>
> >
> > Added a tracepoint to see iostat of f2fs. Default period of that
> > is 3 second. This tracepoint can be used to be monitoring
> > I/O statistics periodically.
> >
> > Bug: 152162885
> > Change-Id: I6fbe010b9cf1a90caa0f4793a6dab77c4cba7da6
>
> It needs to be removed?

Yup.

>
> > Signed-off-by: Daeho Jeong <daehojeong@xxxxxxxxxx>
> > ---
> > fs/f2fs/f2fs.h | 10 ++++++-
> > fs/f2fs/sysfs.c | 34 ++++++++++++++++++++++++
> > include/trace/events/f2fs.h | 52 +++++++++++++++++++++++++++++++++++++
> > 3 files changed, 95 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > index c2788738aa0d4..87baa09f76fb2 100644
> > --- a/fs/f2fs/f2fs.h
> > +++ b/fs/f2fs/f2fs.h
> > @@ -2999,16 +2999,22 @@ static inline int get_inline_xattr_addrs(struct inode *inode)
> > sizeof((f2fs_inode)->field)) \
> > <= (F2FS_OLD_ATTRIBUTE_SIZE + (extra_isize))) \
> >
> > +extern unsigned long long f2fs_prev_iostat[NR_IO_TYPE];
> > +
> > static inline void f2fs_reset_iostat(struct f2fs_sb_info *sbi)
> > {
> > int i;
> >
> > spin_lock(&sbi->iostat_lock);
> > - for (i = 0; i < NR_IO_TYPE; i++)
> > + for (i = 0; i < NR_IO_TYPE; i++) {
> > sbi->write_iostat[i] = 0;
> > + f2fs_prev_iostat[i] = 0;
> > + }
> > spin_unlock(&sbi->iostat_lock);
> > }
> >
> > +extern void f2fs_record_iostat(struct f2fs_sb_info *sbi);
> > +
> > static inline void f2fs_update_iostat(struct f2fs_sb_info *sbi,
> > enum iostat_type type, unsigned long long io_bytes)
> > {
> > @@ -3022,6 +3028,8 @@ static inline void f2fs_update_iostat(struct f2fs_sb_info *sbi,
> > sbi->write_iostat[APP_WRITE_IO] -
> > sbi->write_iostat[APP_DIRECT_IO];
> > spin_unlock(&sbi->iostat_lock);
> > +
> > + f2fs_record_iostat(sbi);
> > }
> >
> > #define __is_large_section(sbi) ((sbi)->segs_per_sec > 1)
> > diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
> > index aeebfb5024a22..f34cb75cd039c 100644
> > --- a/fs/f2fs/sysfs.c
> > +++ b/fs/f2fs/sysfs.c
> > @@ -15,6 +15,7 @@
> > #include "f2fs.h"
> > #include "segment.h"
> > #include "gc.h"
> > +#include <trace/events/f2fs.h>
> >
> > static struct proc_dir_entry *f2fs_proc_root;
> >
> > @@ -751,6 +752,39 @@ static int __maybe_unused segment_bits_seq_show(struct seq_file *seq,
> > return 0;
> > }
> >
> > +static const unsigned long period_ms = 3000;
> > +static unsigned long next_period;
> > +unsigned long long f2fs_prev_iostat[NR_IO_TYPE] = {0};
>
> These various should be per sbi, otherwise stats from different image could
> be interrupt others'.

Agreed. Let me pass it to work on it.

Thanks,

>
> Thanks,
>
> > +
> > +static DEFINE_SPINLOCK(iostat_lock);
> > +
> > +void f2fs_record_iostat(struct f2fs_sb_info *sbi)
> > +{
> > + unsigned long long iostat_diff[NR_IO_TYPE];
> > + int i;
> > +
> > + if (time_is_after_jiffies(next_period))
> > + return;
> > +
> > + /* Need double check under the lock */
> > + spin_lock(&iostat_lock);
> > + if (time_is_after_jiffies(next_period)) {
> > + spin_unlock(&iostat_lock);
> > + return;
> > + }
> > + next_period = jiffies + msecs_to_jiffies(period_ms);
> > + spin_unlock(&iostat_lock);
> > +
> > + spin_lock(&sbi->iostat_lock);
> > + for (i = 0; i < NR_IO_TYPE; i++) {
> > + iostat_diff[i] = sbi->write_iostat[i] - f2fs_prev_iostat[i];
> > + f2fs_prev_iostat[i] = sbi->write_iostat[i];
> > + }
> > + spin_unlock(&sbi->iostat_lock);
> > +
> > + trace_f2fs_iostat(sbi, iostat_diff);
> > +}
> > +
> > static int __maybe_unused iostat_info_seq_show(struct seq_file *seq,
> > void *offset)
> > {
> > diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h
> > index d97adfc327f03..e78c8696e2adc 100644
> > --- a/include/trace/events/f2fs.h
> > +++ b/include/trace/events/f2fs.h
> > @@ -1812,6 +1812,58 @@ DEFINE_EVENT(f2fs_zip_end, f2fs_decompress_pages_end,
> > TP_ARGS(inode, cluster_idx, compressed_size, ret)
> > );
> >
> > +TRACE_EVENT(f2fs_iostat,
> > +
> > + TP_PROTO(struct f2fs_sb_info *sbi, unsigned long long *iostat),
> > +
> > + TP_ARGS(sbi, iostat),
> > +
> > + TP_STRUCT__entry(
> > + __field(dev_t, dev)
> > + __field(unsigned long long, app_dio)
> > + __field(unsigned long long, app_bio)
> > + __field(unsigned long long, app_wio)
> > + __field(unsigned long long, app_mio)
> > + __field(unsigned long long, fs_dio)
> > + __field(unsigned long long, fs_nio)
> > + __field(unsigned long long, fs_mio)
> > + __field(unsigned long long, fs_gc_dio)
> > + __field(unsigned long long, fs_gc_nio)
> > + __field(unsigned long long, fs_cp_dio)
> > + __field(unsigned long long, fs_cp_nio)
> > + __field(unsigned long long, fs_cp_mio)
> > + __field(unsigned long long, fs_discard)
> > + ),
> > +
> > + TP_fast_assign(
> > + __entry->dev = sbi->sb->s_dev;
> > + __entry->app_dio = iostat[APP_DIRECT_IO];
> > + __entry->app_bio = iostat[APP_BUFFERED_IO];
> > + __entry->app_wio = iostat[APP_WRITE_IO];
> > + __entry->app_mio = iostat[APP_MAPPED_IO];
> > + __entry->fs_dio = iostat[FS_DATA_IO];
> > + __entry->fs_nio = iostat[FS_NODE_IO];
> > + __entry->fs_mio = iostat[FS_META_IO];
> > + __entry->fs_gc_dio = iostat[FS_GC_DATA_IO];
> > + __entry->fs_gc_nio = iostat[FS_GC_NODE_IO];
> > + __entry->fs_cp_dio = iostat[FS_CP_DATA_IO];
> > + __entry->fs_cp_nio = iostat[FS_CP_NODE_IO];
> > + __entry->fs_cp_mio = iostat[FS_CP_META_IO];
> > + __entry->fs_discard = iostat[FS_DISCARD];
> > + ),
> > +
> > + TP_printk("dev = (%d,%d), "
> > + "app [write=%llu (direct=%llu, buffered=%llu), mapped=%llu], "
> > + "fs [data=%llu, node=%llu, meta=%llu, discard=%llu], "
> > + "gc [data=%llu, node=%llu], "
> > + "cp [data=%llu, node=%llu, meta=%llu]",
> > + show_dev(__entry->dev), __entry->app_wio, __entry->app_dio,
> > + __entry->app_bio, __entry->app_mio, __entry->fs_dio,
> > + __entry->fs_nio, __entry->fs_mio, __entry->fs_discard,
> > + __entry->fs_gc_dio, __entry->fs_gc_nio, __entry->fs_cp_dio,
> > + __entry->fs_cp_nio, __entry->fs_cp_mio)
> > +);
> > +
> > #endif /* _TRACE_F2FS_H */
> >
> > /* This part must be outside protection */
> >