[PATCH v2] f2fs: get victim segment again after new cp

[PATCH v2] f2fs: get victim segment again after new cp

[Yunlei He]

Previous selected segment may become free after write_checkpoint,
if we do garbage collect on this segment, and then new_curseg happen
to reuse it, it may cause f2fs_bug_on as below.

	panic+0x154/0x29c
	do_garbage_collect+0x15c/0xaf4
	f2fs_gc+0x2dc/0x444
	f2fs_balance_fs.part.22+0xcc/0x14c
	f2fs_balance_fs+0x28/0x34
	f2fs_map_blocks+0x5ec/0x790
	f2fs_preallocate_blocks+0xe0/0x100
	f2fs_file_write_iter+0x64/0x11c
	new_sync_write+0xac/0x11c
	vfs_write+0x144/0x1e4
	SyS_write+0x60/0xc0

Here, maybe we check sit and ssa type during reset_curseg. So, we check
segment is stale or not, and select a new victim to avoid this.

Signed-off-by: Yunlei He <heyunlei@huawei.com>
---
 fs/f2fs/gc.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index de6c41c..ec17096 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -908,10 +908,14 @@ gc_more:
 		 * enough free sections, we should flush dent/node blocks and do
 		 * garbage collections.
 		 */
-		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi))
+		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi)) {
 			write_checkpoint(sbi, &cpc);
-		else if (has_not_enough_free_secs(sbi, 0))
+			if(!get_valid_blocks(sbi, segno, sbi->segs_per_sec)
+					|| IS_CURSEC(sbi, segno / sbi->segs_per_sec))
+				segno = NULL_SEGNO;
+		} else if (has_not_enough_free_secs(sbi, 0)) {
 			write_checkpoint(sbi, &cpc);
+		}
 	}
 
 	if (segno == NULL_SEGNO && !__get_victim(sbi, &segno, gc_type))
-- 
1.9.1


------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are 
consuming the most bandwidth. Provides multi-vendor support for NetFlow, 
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports.http://sdm.link/zohodev2dev

Re: [PATCH v2] f2fs: get victim segment again after new cp

[Jaegeuk Kim]

Hi Yunlei,

On Fri, Jul 22, 2016 at 07:08:31PM +0800, Yunlei He wrote:
> Previous selected segment may become free after write_checkpoint,
> if we do garbage collect on this segment, and then new_curseg happen
> to reuse it, it may cause f2fs_bug_on as below.
> 
> 	panic+0x154/0x29c
> 	do_garbage_collect+0x15c/0xaf4
> 	f2fs_gc+0x2dc/0x444
> 	f2fs_balance_fs.part.22+0xcc/0x14c
> 	f2fs_balance_fs+0x28/0x34
> 	f2fs_map_blocks+0x5ec/0x790
> 	f2fs_preallocate_blocks+0xe0/0x100
> 	f2fs_file_write_iter+0x64/0x11c
> 	new_sync_write+0xac/0x11c
> 	vfs_write+0x144/0x1e4
> 	SyS_write+0x60/0xc0
> 
> Here, maybe we check sit and ssa type during reset_curseg. So, we check
> segment is stale or not, and select a new victim to avoid this.
> 
> Signed-off-by: Yunlei He <heyunlei@huawei.com>
> ---
>  fs/f2fs/gc.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> index de6c41c..ec17096 100644
> --- a/fs/f2fs/gc.c
> +++ b/fs/f2fs/gc.c
> @@ -908,10 +908,14 @@ gc_more:
>  		 * enough free sections, we should flush dent/node blocks and do
>  		 * garbage collections.
>  		 */
> -		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi))
> +		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi)) {
>  			write_checkpoint(sbi, &cpc);
> -		else if (has_not_enough_free_secs(sbi, 0))

If segno is NULL_SEGNO, we get a panic when checking the below conditions.
Anyway, I don't think we need this condition at all.
Let me remove them and just set NULL_SEGNO only after checkpoint.
Please check the dev-test repo.

Thanks,

> +			if(!get_valid_blocks(sbi, segno, sbi->segs_per_sec)
> +					|| IS_CURSEC(sbi, segno / sbi->segs_per_sec))
> +				segno = NULL_SEGNO;
> +		} else if (has_not_enough_free_secs(sbi, 0)) {
>  			write_checkpoint(sbi, &cpc);
> +		}
>  	}
>  
>  	if (segno == NULL_SEGNO && !__get_victim(sbi, &segno, gc_type))
> -- 
> 1.9.1

------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are 
consuming the most bandwidth. Provides multi-vendor support for NetFlow, 
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports.http://sdm.link/zohodev2dev

Re: [PATCH v2] f2fs: get victim segment again after new cp

[Chao Yu]

Hi Jaegeuk,

On 2016/7/23 2:59, Jaegeuk Kim wrote:
> Hi Yunlei,
> 
> On Fri, Jul 22, 2016 at 07:08:31PM +0800, Yunlei He wrote:
>> Previous selected segment may become free after write_checkpoint,
>> if we do garbage collect on this segment, and then new_curseg happen
>> to reuse it, it may cause f2fs_bug_on as below.
>>
>> 	panic+0x154/0x29c
>> 	do_garbage_collect+0x15c/0xaf4
>> 	f2fs_gc+0x2dc/0x444
>> 	f2fs_balance_fs.part.22+0xcc/0x14c
>> 	f2fs_balance_fs+0x28/0x34
>> 	f2fs_map_blocks+0x5ec/0x790
>> 	f2fs_preallocate_blocks+0xe0/0x100
>> 	f2fs_file_write_iter+0x64/0x11c
>> 	new_sync_write+0xac/0x11c
>> 	vfs_write+0x144/0x1e4
>> 	SyS_write+0x60/0xc0
>>
>> Here, maybe we check sit and ssa type during reset_curseg. So, we check
>> segment is stale or not, and select a new victim to avoid this.
>>
>> Signed-off-by: Yunlei He <heyunlei@huawei.com>
>> ---
>>  fs/f2fs/gc.c | 8 ++++++--
>>  1 file changed, 6 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
>> index de6c41c..ec17096 100644
>> --- a/fs/f2fs/gc.c
>> +++ b/fs/f2fs/gc.c
>> @@ -908,10 +908,14 @@ gc_more:
>>  		 * enough free sections, we should flush dent/node blocks and do
>>  		 * garbage collections.
>>  		 */
>> -		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi))
>> +		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi)) {
>>  			write_checkpoint(sbi, &cpc);
>> -		else if (has_not_enough_free_secs(sbi, 0))
> 
> If segno is NULL_SEGNO, we get a panic when checking the below conditions.
> Anyway, I don't think we need this condition at all.

I thinks the condition which trigger our f2fs_bug_on is really a corner case,
we'd better not waste the victim we found everytime, especially for these
victims from last background gc's victims.

What about checking segno first?

if(segno != NULL_SEGNO && (!get_valid_blocks(sbi, segno, sbi->segs_per_sec) ||
	IS_CURSEC(sbi, GET_SECNO(sbi, segno)))
	segno = NULL_SEGNO;

Thanks,

> Let me remove them and just set NULL_SEGNO only after checkpoint.
> Please check the dev-test repo.
> 
> Thanks,
> 
>> +			if(!get_valid_blocks(sbi, segno, sbi->segs_per_sec)
>> +					|| IS_CURSEC(sbi, segno / sbi->segs_per_sec))
>> +				segno = NULL_SEGNO;
>> +		} else if (has_not_enough_free_secs(sbi, 0)) {
>>  			write_checkpoint(sbi, &cpc);
>> +		}
>>  	}
>>  
>>  	if (segno == NULL_SEGNO && !__get_victim(sbi, &segno, gc_type))
>> -- 
>> 1.9.1
> 
> ------------------------------------------------------------------------------
> What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
> patterns at an interface-level. Reveals which users, apps, and protocols are 
> consuming the most bandwidth. Provides multi-vendor support for NetFlow, 
> J-Flow, sFlow and other flows. Make informed decisions using capacity planning
> reports.http://sdm.link/zohodev2dev
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
> 

------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are 
consuming the most bandwidth. Provides multi-vendor support for NetFlow, 
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports.http://sdm.link/zohodev2dev

Re: [PATCH v2] f2fs: get victim segment again after new cp

[Jaegeuk Kim]

On Sun, Jul 24, 2016 at 06:38:46AM +0800, Chao Yu wrote:
...

> >> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> >> index de6c41c..ec17096 100644
> >> --- a/fs/f2fs/gc.c
> >> +++ b/fs/f2fs/gc.c
> >> @@ -908,10 +908,14 @@ gc_more:
> >>  		 * enough free sections, we should flush dent/node blocks and do
> >>  		 * garbage collections.
> >>  		 */
> >> -		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi))
> >> +		if (__get_victim(sbi, &segno, gc_type) || prefree_segments(sbi)) {
> >>  			write_checkpoint(sbi, &cpc);
> >> -		else if (has_not_enough_free_secs(sbi, 0))
> > 
> > If segno is NULL_SEGNO, we get a panic when checking the below conditions.
> > Anyway, I don't think we need this condition at all.
> 
> I thinks the condition which trigger our f2fs_bug_on is really a corner case,
> we'd better not waste the victim we found everytime, especially for these
> victims from last background gc's victims.
> 
> What about checking segno first?
> 
> if(segno != NULL_SEGNO && (!get_valid_blocks(sbi, segno, sbi->segs_per_sec) ||
> 	IS_CURSEC(sbi, GET_SECNO(sbi, segno)))
> 	segno = NULL_SEGNO;

My viewpoint is write_checkpoint() flushes node and dentry pages, which provides
another better victims. Another concern is get_valid_blocks and IS_CURSEG are
not covered by any lock.

Thanks,

> 
> Thanks,
> 
> > Let me remove them and just set NULL_SEGNO only after checkpoint.
> > Please check the dev-test repo.
> > 
> > Thanks,
> > 
> >> +			if(!get_valid_blocks(sbi, segno, sbi->segs_per_sec)
> >> +					|| IS_CURSEC(sbi, segno / sbi->segs_per_sec))
> >> +				segno = NULL_SEGNO;
> >> +		} else if (has_not_enough_free_secs(sbi, 0)) {
> >>  			write_checkpoint(sbi, &cpc);
> >> +		}
> >>  	}
> >>  
> >>  	if (segno == NULL_SEGNO && !__get_victim(sbi, &segno, gc_type))
> >> -- 
> >> 1.9.1
> > 
> > ------------------------------------------------------------------------------
> > What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
> > patterns at an interface-level. Reveals which users, apps, and protocols are 
> > consuming the most bandwidth. Provides multi-vendor support for NetFlow, 
> > J-Flow, sFlow and other flows. Make informed decisions using capacity planning
> > reports.http://sdm.link/zohodev2dev
> > _______________________________________________
> > Linux-f2fs-devel mailing list
> > Linux-f2fs-devel@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
> > 

------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are 
consuming the most bandwidth. Provides multi-vendor support for NetFlow, 
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports.http://sdm.link/zohodev2dev