2013-04-25 02:45:09

by Jaegeuk Kim

[permalink] [raw]
Subject: [PATCH] f2fs: avoid frequent background GC

If there is no victim segments selected by background GC, let's wait
a little bit longer time to collect dirty segments.
By default, let's give 10 minutes.

Signed-off-by: Jaegeuk Kim <[email protected]>
---
fs/f2fs/gc.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/gc.h b/fs/f2fs/gc.h
index 30b2db0..1190d57 100644
--- a/fs/f2fs/gc.h
+++ b/fs/f2fs/gc.h
@@ -13,9 +13,9 @@
* whether IO subsystem is idle
* or not
*/
-#define GC_THREAD_MIN_SLEEP_TIME 10000 /* milliseconds */
+#define GC_THREAD_MIN_SLEEP_TIME 10000 /* milliseconds */
#define GC_THREAD_MAX_SLEEP_TIME 30000
-#define GC_THREAD_NOGC_SLEEP_TIME 10000
+#define GC_THREAD_NOGC_SLEEP_TIME 600000 /* wait 10 min */
#define LIMIT_INVALID_BLOCK 40 /* percentage over total user space */
#define LIMIT_FREE_BLOCK 40 /* percentage over invalid + free space */

--
1.8.1.3.566.gaa39828


2013-04-25 05:06:48

by Namjae Jeon

[permalink] [raw]
Subject: Re: [PATCH] f2fs: avoid frequent background GC

Hi Jaegeuk,

With the change in the value of GC_THREAD_NOGC_SLEEP_TIME to 600000,
we will need to rewrite the gc_thread_func().
As there will be several paths which will not be reached or will have
no meaning after this change.

Considering the cases:
/* if return value is not zero, no victim was selected */
if (f2fs_gc(sbi))
wait_ms = GC_THREAD_NOGC_SLEEP_TIME;
else if (wait_ms == GC_THREAD_NOGC_SLEEP_TIME)
wait_ms = GC_THREAD_MAX_SLEEP_TIME;

Before this point, there are ‘2’ functions to change ‘wait_ms” ->
increase_sleep_time and decrease_sleep_time.
When we consider increase_sleep_time() - the maximum possible value
for ‘wait_ms’ after coming out of this can be:
GC_THREAD_MAX_SLEEP_TIME
And in case in the last iteration it was set to
GC_THREAD_NOGC_SLEEP_TIME, once it enters increase_sleep_time() - the
value will be reseted to maximum GC_THREAD_MAX_SLEEP_TIME.
So,
else if (wait_ms == GC_THREAD_NOGC_SLEEP_TIME)
wait_ms = GC_THREAD_MAX_SLEEP_TIME;
becomes unreachable path.

And second, maybe, After setting to GC_THREAD_NOGC_SLEEP_TIME(10min),
Although there are invalid blocks, GC thread might be sleep 9min
50sec.
Let me know your opinion.

Thanks.

2013/4/25, Jaegeuk Kim <[email protected]>:
> If there is no victim segments selected by background GC, let's wait
> a little bit longer time to collect dirty segments.
> By default, let's give 10 minutes.
>
> Signed-off-by: Jaegeuk Kim <[email protected]>
> ---
> fs/f2fs/gc.h | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/fs/f2fs/gc.h b/fs/f2fs/gc.h
> index 30b2db0..1190d57 100644
> --- a/fs/f2fs/gc.h
> +++ b/fs/f2fs/gc.h
> @@ -13,9 +13,9 @@
> * whether IO subsystem is idle
> * or not
> */
> -#define GC_THREAD_MIN_SLEEP_TIME 10000 /* milliseconds */
> +#define GC_THREAD_MIN_SLEEP_TIME 10000 /* milliseconds */
> #define GC_THREAD_MAX_SLEEP_TIME 30000
> -#define GC_THREAD_NOGC_SLEEP_TIME 10000
> +#define GC_THREAD_NOGC_SLEEP_TIME 600000 /* wait 10 min */
> #define LIMIT_INVALID_BLOCK 40 /* percentage over total user space */
> #define LIMIT_FREE_BLOCK 40 /* percentage over invalid + free space */
>
> --
> 1.8.1.3.566.gaa39828
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>

2013-04-25 11:01:53

by Jaegeuk Kim

[permalink] [raw]
Subject: Re: [PATCH v2] f2fs: avoid frequent background GC

Hi, Namjae,

Agreed. How about this?

Chang log from v1:
o change timings - min 30s, max 60s, nogc 5 min
o remove nonreachable routine
o consider NOGC_SLEEP_TIME in increate/decrease_sleep_time

From 806e344624414fcf9fc87f6193265859027d51b5 Mon Sep 17 00:00:00 2001
From: Jaegeuk Kim <[email protected]>
Date: Wed, 24 Apr 2013 13:00:14 +0900
Subject: [PATCH] f2fs: avoid frequent background GC
Cc: [email protected], [email protected],
[email protected]

If there is no victim segments selected by background GC, let's wait
a little bit longer time to collect dirty segments.
By default, let's give 5 minutes.

Signed-off-by: Jaegeuk Kim <[email protected]>
---
fs/f2fs/gc.c | 3 ---
fs/f2fs/gc.h | 12 +++++++++---
2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 1ca3324..6ed3263 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -82,9 +82,6 @@ static int gc_thread_func(void *data)
/* if return value is not zero, no victim was selected */
if (f2fs_gc(sbi))
wait_ms = GC_THREAD_NOGC_SLEEP_TIME;
- else if (wait_ms == GC_THREAD_NOGC_SLEEP_TIME)
- wait_ms = GC_THREAD_MAX_SLEEP_TIME;
-
} while (!kthread_should_stop());
return 0;
}
diff --git a/fs/f2fs/gc.h b/fs/f2fs/gc.h
index 30b2db0..2c6a6bd 100644
--- a/fs/f2fs/gc.h
+++ b/fs/f2fs/gc.h
@@ -13,9 +13,9 @@
* whether IO subsystem is idle
* or not
*/
-#define GC_THREAD_MIN_SLEEP_TIME 10000 /* milliseconds */
-#define GC_THREAD_MAX_SLEEP_TIME 30000
-#define GC_THREAD_NOGC_SLEEP_TIME 10000
+#define GC_THREAD_MIN_SLEEP_TIME 30000 /* milliseconds */
+#define GC_THREAD_MAX_SLEEP_TIME 60000
+#define GC_THREAD_NOGC_SLEEP_TIME 300000 /* wait 5 min */
#define LIMIT_INVALID_BLOCK 40 /* percentage over total user space */
#define LIMIT_FREE_BLOCK 40 /* percentage over invalid + free space */

@@ -58,6 +58,9 @@ static inline block_t limit_free_user_blocks(struct
f2fs_sb_info *sbi)

static inline long increase_sleep_time(long wait)
{
+ if (wait == GC_THREAD_NOGC_SLEEP_TIME)
+ return wait;
+
wait += GC_THREAD_MIN_SLEEP_TIME;
if (wait > GC_THREAD_MAX_SLEEP_TIME)
wait = GC_THREAD_MAX_SLEEP_TIME;
@@ -66,6 +69,9 @@ static inline long increase_sleep_time(long wait)

static inline long decrease_sleep_time(long wait)
{
+ if (wait == GC_THREAD_NOGC_SLEEP_TIME)
+ wait = GC_THREAD_MAX_SLEEP_TIME;
+
wait -= GC_THREAD_MIN_SLEEP_TIME;
if (wait <= GC_THREAD_MIN_SLEEP_TIME)
wait = GC_THREAD_MIN_SLEEP_TIME;
--
1.8.1.3.566.gaa39828



--
Jaegeuk Kim
Samsung


Attachments:
signature.asc (836.00 B)
This is a digitally signed message part

2013-04-25 23:52:42

by Namjae Jeon

[permalink] [raw]
Subject: Re: [PATCH v2] f2fs: avoid frequent background GC

2013/4/25, Jaegeuk Kim <[email protected]>:
> Hi, Namjae,
>
> Agreed. How about this?
>
> Chang log from v1:
> o change timings - min 30s, max 60s, nogc 5 min
> o remove nonreachable routine
> o consider NOGC_SLEEP_TIME in increate/decrease_sleep_time
>
> From 806e344624414fcf9fc87f6193265859027d51b5 Mon Sep 17 00:00:00 2001
> From: Jaegeuk Kim <[email protected]>
> Date: Wed, 24 Apr 2013 13:00:14 +0900
> Subject: [PATCH] f2fs: avoid frequent background GC
> Cc: [email protected], [email protected],
> linu[email protected]
>
> If there is no victim segments selected by background GC, let's wait
> a little bit longer time to collect dirty segments.
> By default, let's give 5 minutes.
>
> Signed-off-by: Jaegeuk Kim <[email protected]>
Looks good!
Reviewed-by: Namjae Jeon <[email protected]>

Thanks~