[Cluster-devel] [PATCH] sub sd_rgrps When clear rgrp

[Cluster-devel] [PATCH] sub sd_rgrps When clear rgrp

[Yang Bin]

From: " Yang Bin "<yang.bin18@zte.com.cn>

When clear rgrp,sub sd_rgrps after erased from rindex_tree

Signed-off-by: Yang Bin <yang.bin18@zte.com.cn>
---
 fs/gfs2/rgrp.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 15d6e32..a4b2e83
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -730,6 +730,7 @@ void gfs2_clear_rgrpd(struct gfs2_sbd *sdp)
 		gl = rgd->rd_gl;
 
 		rb_erase(n, &sdp->sd_rindex_tree);
+		sdp->sd_rgrps--;
 
 		if (gl) {
 			glock_clear_object(gl, rgd);
-- 
1.8.3.1

[Cluster-devel] [PATCH] sub sd_rgrps When clear rgrp

[Andreas Grünbacher]

Hi Yang Bin,

Am Do., 27. Juni 2019 um 05:08 Uhr schrieb Yang Bin <yang.bin18@zte.com.cn>:
> From: " Yang Bin "<yang.bin18@zte.com.cn>
>
> When clear rgrp,sub sd_rgrps after erased from rindex_tree

this patch isn't incorrect, but all it does it ensure that sd_rgrps
reaches zero before we destroy a struct gfs2_sbd. However, sd_rgrps is
only ever used when a filesystem is active, and while that is the
case, it can never decrease. It will grow when the filesystem is grown
though.

So what are you trying to achieve with this patch?

> Signed-off-by: Yang Bin <yang.bin18@zte.com.cn>
> ---
>  fs/gfs2/rgrp.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
> index 15d6e32..a4b2e83
> --- a/fs/gfs2/rgrp.c
> +++ b/fs/gfs2/rgrp.c
> @@ -730,6 +730,7 @@ void gfs2_clear_rgrpd(struct gfs2_sbd *sdp)
>                 gl = rgd->rd_gl;
>
>                 rb_erase(n, &sdp->sd_rindex_tree);
> +               sdp->sd_rgrps--;
>
>                 if (gl) {
>                         glock_clear_object(gl, rgd);
> --
> 1.8.3.1
>

Thanks,
Andreas

[Cluster-devel] [PATCH] sub sd_rgrps When clear rgrp

[yang.bin18]

Hi Andreas,

>Am Do., 27. Juni 2019 um 05:08 Uhr schrieb Yang Bin <yang.bin18@zte.com.cn>:
>> From: " Yang Bin "<yang.bin18@zte.com.cn>
>>
>> When clear rgrp,sub sd_rgrps after erased from rindex_tree

>this patch isn't incorrect, but all it does it ensure that sd_rgrps
>reaches zero before we destroy a struct gfs2_sbd. However, sd_rgrps is
>only ever used when a filesystem is active, and while that is the
>case, it can never decrease. It will grow when the filesystem is grown
>though.

>So what are you trying to achieve with this patch?
In mounting, it may fail and then execute "gfs2_clear_rgrpd" . To solve this failing ,
I add reload process,which will read rindex again .But reload will  fail ,
because sd_rgrps has beyonded rindex`s size. So I make this patch.

>> Signed-off-by: Yang Bin <yang.bin18@zte.com.cn>
>> ---
>>  fs/gfs2/rgrp.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
>> index 15d6e32..a4b2e83
>> --- a/fs/gfs2/rgrp.c
>> +++ b/fs/gfs2/rgrp.c
>> @@ -730,6 +730,7 @@ void gfs2_clear_rgrpd(struct gfs2_sbd *sdp)
>>                 gl = rgd->rd_gl;
>>
>>                 rb_erase(n, &sdp->sd_rindex_tree);
>> +               sdp->sd_rgrps--;
>>
>>                 if (gl) {
>>                         glock_clear_object(gl, rgd);
>> --
>> 1.8.3.1
>>
>Thanks,
>Andreas