[PATCHv2] nilfs2: shorten freeze period due to GC in write operation

[PATCHv2] nilfs2: shorten freeze period due to GC in write operation

[Jiro SEKIBA]

Hi,

This is a revised patch to shorten freeze period.
This version includes fixes of the bugs Konishi-san mentioned.

When GC is runnning, GC moves live block to difference segments.
Copying live blocks into memory is done in a transaction,
but it is not necessarily to be in the transaction.
This patch will get the nilfs_ioctl_move_blocks() out from
transaction lock and put it before the transaction.

I ran sysbench fileio test against nilfs partition.
I copied some DVD/CD images and created snapshot to create live blocks
before starting the benchmark.

Followings are summary of rc8 and rc8 w/ the patch of per-request statistics,
which is min/max and avg.  I ran each test three times and bellow is
average of those numers.

According to this benchmark result, average time is slightly degrated.
However, worstcase (max) result is significantly improved.
This can address a few seconds write freeze.

- random write per-request performance of rc8
 min   0.843ms
 max 680.406ms
 avg   3.050ms
- random write per-request performance of rc8 w/ this patch
 min   0.830ms ->  98.40%
 max 366.870ms ->  53.90%
 avg   3.170ms -> 103.90%

- sequential write per-request performance of rc8
 min   0.736ms
 max 774.343ms 
 avg   2.883ms 
- sequential write per-request performance of rc8 w/ this patch
 min   0.726ms -> 100.50%
 max  695.976ms->  72.80%
 avg   3.110ms -> 107.80%

-----8<-----8<-----nilfs_cleanerd.conf-----8<-----8<-----
protection_period       150
selection_policy        timestamp       # timestamp in ascend order
nsegments_per_clean     2
cleaning_interval       2
retry_interval          60
use_mmap
log_priority            info
-----8<-----8<-----nilfs_cleanerd.conf-----8<-----8<-----


Signed-off-by: Jiro SEKIBA <jir-hfpbi5WX9J54Eiagz67IpQ@public.gmane.org>
---
 fs/nilfs2/ioctl.c     |   22 ++++++++++++++--------
 fs/nilfs2/the_nilfs.h |    2 ++
 2 files changed, 16 insertions(+), 8 deletions(-)

diff --git a/fs/nilfs2/ioctl.c b/fs/nilfs2/ioctl.c
index 6ea5f87..cc6ec4b 100644
--- a/fs/nilfs2/ioctl.c
+++ b/fs/nilfs2/ioctl.c
@@ -442,12 +442,6 @@ int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs,
 	const char *msg;
 	int ret;
 
-	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
-	if (ret < 0) {
-		msg = "cannot read source blocks";
-		goto failed;
-	}
-
 	ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]);
 	if (ret < 0) {
 		/*
@@ -509,6 +503,12 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 	nsegs = argv[4].v_nmembs;
 	if (argv[4].v_size != argsz[4])
 		return -EINVAL;
+
+	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
+
+	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags))
+		return -EBUSY;
+
 	/*
 	 * argv[4] points to segment numbers this ioctl cleans.  We
 	 * use kmalloc() for its buffer because memory used for the
@@ -519,8 +519,6 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 	if (IS_ERR(kbufs[4]))
 		return PTR_ERR(kbufs[4]);
 
-	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
-
 	for (n = 0; n < 4; n++) {
 		ret = -EINVAL;
 		if (argv[n].v_size != argsz[n])
@@ -548,12 +546,20 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 		}
 	}
 
+	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
+	if (ret < 0) {
+		printk(KERN_ERR "NILFS: GC failed during preparation: "
+			"cannot read source blocks: err=%d\n", ret);
+		goto out_free;
+	}
+
 	ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
 
  out_free:
 	while (--n >= 0)
 		vfree(kbufs[n]);
 	kfree(kbufs[4]);
+	clear_nilfs_gc_running(nilfs);
 	return ret;
 }
 
diff --git a/fs/nilfs2/the_nilfs.h b/fs/nilfs2/the_nilfs.h
index 1b9caaf..97ee569 100644
--- a/fs/nilfs2/the_nilfs.h
+++ b/fs/nilfs2/the_nilfs.h
@@ -37,6 +37,7 @@ enum {
 	THE_NILFS_LOADED,       /* Roll-back/roll-forward has done and
 				   the latest checkpoint was loaded */
 	THE_NILFS_DISCONTINUED,	/* 'next' pointer chain has broken */
+	THE_NILFS_GC_RUNNING,	/* gc process is running */
 };
 
 /**
@@ -197,6 +198,7 @@ static inline int nilfs_##name(struct the_nilfs *nilfs)			\
 THE_NILFS_FNS(INIT, init)
 THE_NILFS_FNS(LOADED, loaded)
 THE_NILFS_FNS(DISCONTINUED, discontinued)
+THE_NILFS_FNS(GC_RUNNING, gc_running)
 
 /* Minimum interval of periodical update of superblocks (in seconds) */
 #define NILFS_SB_FREQ		10
-- 
1.5.6.5

Re: [PATCHv2] nilfs2: shorten freeze period due to GC in write operation

[Ryusuke Konishi]

Hi Sekiba-san,
On Wed,  2 Sep 2009 15:29:50 +0900, Jiro SEKIBA <jir-hfpbi5WX9J54Eiagz67IpQ@public.gmane.org> wrote:
> Hi,
> 
> This is a revised patch to shorten freeze period.
> This version includes fixes of the bugs Konishi-san mentioned.
> <snip>
> @@ -509,6 +503,12 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
>  	nsegs = argv[4].v_nmembs;
>  	if (argv[4].v_size != argsz[4])
>  		return -EINVAL;
> +
> +	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
> +
> +	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags))
> +		return -EBUSY;
> +
>  	/*
>  	 * argv[4] points to segment numbers this ioctl cleans.  We
>  	 * use kmalloc() for its buffer because memory used for the
> @@ -519,8 +519,6 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
>  	if (IS_ERR(kbufs[4]))
>  		return PTR_ERR(kbufs[4]);

Hmmm, moving forward the test-and-set operation was not good idea.
Here is another problem.  When memory allocation fails, the process
will return this function without clearing the GC bit.

How about putting bit operations adjacent to the GC routines as
follows ?

	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) {
		ret = -EBUSY;
		goto out_free;
	}

	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
	if (ret < 0)
		printk(KERN_ERR "NILFS: GC failed during preparation: "
		       "cannot read source blocks: err=%d\n", ret);
	else
		ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);

	clear_nilfs_gc_running(nilfs);

I think it's also preferable to highlight what we should protect.

Thanks,
Ryusuke Konishi.

Re: [PATCHv2] nilfs2: shorten freeze period due to GC in write operation

[Jiro SEKIBA]

Hi,

At Wed, 02 Sep 2009 22:20:35 +0900 (JST),
Ryusuke Konishi wrote:
> 
> Hi Sekiba-san,
> On Wed,  2 Sep 2009 15:29:50 +0900, Jiro SEKIBA <jir-hfpbi5WX9J54Eiagz67IpQ@public.gmane.org> wrote:
> > Hi,
> > 
> > This is a revised patch to shorten freeze period.
> > This version includes fixes of the bugs Konishi-san mentioned.
> > <snip>
> > @@ -509,6 +503,12 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
> >  	nsegs = argv[4].v_nmembs;
> >  	if (argv[4].v_size != argsz[4])
> >  		return -EINVAL;
> > +
> > +	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
> > +
> > +	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags))
> > +		return -EBUSY;
> > +
> >  	/*
> >  	 * argv[4] points to segment numbers this ioctl cleans.  We
> >  	 * use kmalloc() for its buffer because memory used for the
> > @@ -519,8 +519,6 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
> >  	if (IS_ERR(kbufs[4]))
> >  		return PTR_ERR(kbufs[4]);
> 
> Hmmm, moving forward the test-and-set operation was not good idea.
> Here is another problem.  When memory allocation fails, the process
> will return this function without clearing the GC bit.

That's right.  sorry for the careless misses.
I'll update the patch.

thank you for the review and suggestions

> How about putting bit operations adjacent to the GC routines as
> follows ?
> 
> 	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) {
> 		ret = -EBUSY;
> 		goto out_free;
> 	}
> 
> 	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
> 	if (ret < 0)
> 		printk(KERN_ERR "NILFS: GC failed during preparation: "
> 		       "cannot read source blocks: err=%d\n", ret);
> 	else
> 		ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
> 
> 	clear_nilfs_gc_running(nilfs);
> 
> I think it's also preferable to highlight what we should protect.
> 
> Thanks,
> Ryusuke Konishi.
> _______________________________________________
> users mailing list
> users-JrjvKiOkagjYtjvyW6yDsg@public.gmane.org
> https://www.nilfs.org/mailman/listinfo/users
> 
> 
> 


-- 
Jiro SEKIBA <jir-hfpbi5WX9J54Eiagz67IpQ@public.gmane.org>

Re: [PATCHv2] nilfs2: shorten freeze period due to GC in write operation

[Ryusuke Konishi]

Hi,
On Wed,  2 Sep 2009 15:29:50 +0900, Jiro SEKIBA wrote:
> Hi,
> 
> This is a revised patch to shorten freeze period.
> This version includes fixes of the bugs Konishi-san mentioned.
> 
> When GC is runnning, GC moves live block to difference segments.
> Copying live blocks into memory is done in a transaction,
> but it is not necessarily to be in the transaction.
> This patch will get the nilfs_ioctl_move_blocks() out from
> transaction lock and put it before the transaction.
> 
> I ran sysbench fileio test against nilfs partition.
> I copied some DVD/CD images and created snapshot to create live blocks
> before starting the benchmark.
> 
> Followings are summary of rc8 and rc8 w/ the patch of per-request statistics,
> which is min/max and avg.  I ran each test three times and bellow is
> average of those numers.
> 
> According to this benchmark result, average time is slightly degrated.
> However, worstcase (max) result is significantly improved.
> This can address a few seconds write freeze.
> 
> - random write per-request performance of rc8
>  min   0.843ms
>  max 680.406ms
>  avg   3.050ms
> - random write per-request performance of rc8 w/ this patch
>  min   0.830ms ->  98.40%
>  max 366.870ms ->  53.90%
>  avg   3.170ms -> 103.90%
> 
> - sequential write per-request performance of rc8
>  min   0.736ms
>  max 774.343ms 
>  avg   2.883ms 
> - sequential write per-request performance of rc8 w/ this patch
>  min   0.726ms -> 100.50%
>  max  695.976ms->  72.80%
>  avg   3.110ms -> 107.80%
> 
> -----8<-----8<-----nilfs_cleanerd.conf-----8<-----8<-----
> protection_period       150
> selection_policy        timestamp       # timestamp in ascend order
> nsegments_per_clean     2
> cleaning_interval       2
> retry_interval          60
> use_mmap
> log_priority            info
> -----8<-----8<-----nilfs_cleanerd.conf-----8<-----8<-----
> 
> 
> Signed-off-by: Jiro SEKIBA <jir-hfpbi5WX9J54Eiagz67IpQ@public.gmane.org>
> ---
>  fs/nilfs2/ioctl.c     |   22 ++++++++++++++--------
>  fs/nilfs2/the_nilfs.h |    2 ++
>  2 files changed, 16 insertions(+), 8 deletions(-)
> 
> diff --git a/fs/nilfs2/ioctl.c b/fs/nilfs2/ioctl.c
> index 6ea5f87..cc6ec4b 100644
> --- a/fs/nilfs2/ioctl.c
> +++ b/fs/nilfs2/ioctl.c
> @@ -442,12 +442,6 @@ int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs,
>  	const char *msg;
>  	int ret;
>  
> -	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
> -	if (ret < 0) {
> -		msg = "cannot read source blocks";
> -		goto failed;
> -	}
> -
>  	ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]);
>  	if (ret < 0) {
>  		/*
> @@ -509,6 +503,12 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
>  	nsegs = argv[4].v_nmembs;
>  	if (argv[4].v_size != argsz[4])
>  		return -EINVAL;
> +
> +	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
> +
> +	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags))
> +		return -EBUSY;
> +
>  	/*
>  	 * argv[4] points to segment numbers this ioctl cleans.  We
>  	 * use kmalloc() for its buffer because memory used for the
> @@ -519,8 +519,6 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
>  	if (IS_ERR(kbufs[4]))
>  		return PTR_ERR(kbufs[4]);
>  
> -	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
> -
>  	for (n = 0; n < 4; n++) {
>  		ret = -EINVAL;
>  		if (argv[n].v_size != argsz[n])
> @@ -548,12 +546,20 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
>  		}
>  	}
>  
> +	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
> +	if (ret < 0) {
> +		printk(KERN_ERR "NILFS: GC failed during preparation: "
> +			"cannot read source blocks: err=%d\n", ret);
> +		goto out_free;
> +	}
> +
>  	ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
>  
>   out_free:
>  	while (--n >= 0)
>  		vfree(kbufs[n]);
>  	kfree(kbufs[4]);
> +	clear_nilfs_gc_running(nilfs);
>  	return ret;
>  }
>  
> diff --git a/fs/nilfs2/the_nilfs.h b/fs/nilfs2/the_nilfs.h
> index 1b9caaf..97ee569 100644
> --- a/fs/nilfs2/the_nilfs.h
> +++ b/fs/nilfs2/the_nilfs.h
> @@ -37,6 +37,7 @@ enum {
>  	THE_NILFS_LOADED,       /* Roll-back/roll-forward has done and
>  				   the latest checkpoint was loaded */
>  	THE_NILFS_DISCONTINUED,	/* 'next' pointer chain has broken */
> +	THE_NILFS_GC_RUNNING,	/* gc process is running */
>  };
>  
>  /**
> @@ -197,6 +198,7 @@ static inline int nilfs_##name(struct the_nilfs *nilfs)			\
>  THE_NILFS_FNS(INIT, init)
>  THE_NILFS_FNS(LOADED, loaded)
>  THE_NILFS_FNS(DISCONTINUED, discontinued)
> +THE_NILFS_FNS(GC_RUNNING, gc_running)
>  
>  /* Minimum interval of periodical update of superblocks (in seconds) */
>  #define NILFS_SB_FREQ		10
> -- 
> 1.5.6.5

Thank you very much for the patient effort.

I've just replaced this patch in the -for-next branch.

Thanks,
Ryusuke Konishi