fs/ocfs2: fix race in ocfs2_dentry_attach_lock
PATCH v3 1/1 fs/ocfs2: fix race in ocfs2_dentry_attach_lock
ocfs2_dentry_attach_lock() can be executed in parallel threads against the same dentry. Make that race safe.
The race is like this:
thread A thread B
(A1) enter ocfs2_dentry_attach_lock,
seeing dentry->d_fsdata is NULL,
and no alias found byocfs2_find_local_alias, so kmalloc
a new ocfs2_dentry_lock structure
to local variable "dl", dl1
.....
(B1) enter ocfs2_dentry_attach_lock, seeing dentry->d_fsdata is NULL, and no alias found by ocfs2_find_local_alias so kmalloc a new ocfs2_dentry_lock structure to local variable "dl", dl2.
......
(A2) set dentry->d_fsdata with dl1,
call ocfs2_dentry_lock() and increasedl1->dl_lockres.l_ro_holders to 1 on
success.
......
(B2) set dentry->d_fsdata with dl2
call ocfs2_dentry_lock() and increase
dl2->dl_lockres.l_ro_holders to 1 on
success.
......(A3) call ocfs2_dentry_unlock()
and decreasedl2->dl_lockres.l_ro_holders to 0
on success.
....
(B3) call ocfs2_dentry_unlock(),
decreasing
dl2->dl_lockres.l_ro_holders, but
see it's zero now, panic---
v3: add Reviewed-by, Reported-by and Tested-by only
v2: 1) removed lock on dentry_attach_lock at the first access of
dentry->d_fsdata since it helps very little.
2) do cleanups before freeing the duplicated dl
3) return after freeing the duplicated dl found.
--- fs/ocfs2/dcache.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)
fs/ocfs2/dcache.c +14 -2
@@ -230,6 +230,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,int ret;struct dentry *alias;struct ocfs2_dentry_lock *dl = dentry->d_fsdata; +struct ocfs2_dentry_lock *dl_free_on_race =NULL ;trace_ocfs2_dentry_attach_lock (dentry->d_name.len, dentry->d_name.name, (unsigned long long )parent_blkno, dl);
@@ -310,10 +311,21 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, out_attach:spin_lock (&dentry_attach_lock); - dentry->d_fsdata = dl; - dl->dl_count++; +/* d_fsdata could be set by parallel thread */ +if (unlikely (dentry->d_fsdata && !alias)) { + dl_free_on_race = dl; + }else { + dentry->d_fsdata = dl; + dl->dl_count++; + }spin_unlock (&dentry_attach_lock); +if (unlikely (dl_free_on_race)) { +iput (dl_free_on_race->dl_inode); +ocfs2_lock_res_free (&dl_free_on_race->dl_lockres); +kfree (dl_free_on_race); +return 0 ; + }/* * This actually gets us our PRMODE level lock. From now on, * we'll have a notification if one of these names is
Signature · 1 line
2.13.6
PATCH v3 1/1 fs/ocfs2: fix race in ocfs2_dentry_attach_lock
Hi Wengang,
Quoted 48 lines · Wengang Wang
ocfs2_dentry_attach_lock()can be executed in parallel threads against the same dentry. Make that race safe.
The race is like this:thread A thread B
(A1) enter
ocfs2_dentry_attach_lock,
seeingdentry->d_fsdatais NULL,
and no alias found byocfs2_find_local_alias, so kmalloc
a newocfs2_dentry_lockstructure
to local variable "dl", dl1.....
(B1) enter
ocfs2_dentry_attach_lock, seeingdentry->d_fsdatais NULL, and no alias found byocfs2_find_local_aliasso kmalloc a newocfs2_dentry_lockstructure to local variable "dl", dl2.......
(A2) set
dentry->d_fsdatawith dl1,
callocfs2_dentry_lock()and increasedl1->dl_lockres.l_ro_holdersto 1 on
success....... (B2) set dentry->d_fsdata with dl2 call ocfs2_dentry_lock() and increase dl2->dl_lockres.l_ro_holders to 1 on success. ......(A3) call
ocfs2_dentry_unlock()
and decreasedl2->dl_lockres.l_ro_holdersto 0
on success..... (B3) call ocfs2_dentry_unlock(), decreasing dl2->dl_lockres.l_ro_holders, but see it's zero now, panic
Quoted 4 lines
Signed-off-by Wengang Wang <[email protected]>Reported-by Daniel Sobe <[email protected]>Tested-by Daniel Sobe <[email protected]>Reviewed-by Changwei Ge <[email protected]>
Thanks for the detailed description.
IIUC, you are trying to identify the race case and free the second dentry lock.
So would the following be more clear? Something like:
if (unlikely(dentry->d_fsdata && !alias)) {
dl_attached = true;
} else
...
}...
if (unlikely(dl_attached)) {
ocfs2_lock_res_free(&dl->dl_lockres);
kfree(dl);
iput(inode);
return 0;
}Thanks,
Joseph
Quoted 48 lines
---
v3: add Reviewed-by, Reported-by and Tested-by onlyv2: 1) removed lock on
dentry_attach_lockat the first access of
dentry->d_fsdatasince it helps very little.
2) do cleanups before freeing the duplicated dl
3) return after freeing the duplicated dl found.--- fs/ocfs2/dcache.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)fs/ocfs2/dcache.c +14
-2@@ -230,6 +230,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,int ret;struct dentry *alias;struct ocfs2_dentry_lock *dl = dentry->d_fsdata; +struct ocfs2_dentry_lock *dl_free_on_race =NULL ;trace_ocfs2_dentry_attach_lock (dentry->d_name.len, dentry->d_name.name, (unsigned long long )parent_blkno, dl);@@ -310,10 +311,21 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, out_attach:spin_lock (&dentry_attach_lock); - dentry->d_fsdata = dl; - dl->dl_count++; +/* d_fsdata could be set by parallel thread */ +if (unlikely (dentry->d_fsdata && !alias)) { + dl_free_on_race = dl; + }else { + dentry->d_fsdata = dl; + dl->dl_count++; + }spin_unlock (&dentry_attach_lock); +if (unlikely (dl_free_on_race)) { +iput (dl_free_on_race->dl_inode); +ocfs2_lock_res_free (&dl_free_on_race->dl_lockres); +kfree (dl_free_on_race); +return 0 ; + }/* * This actually gets us our PRMODE level lock. From now on, * we'll have a notification if one of these names is
PATCH v3 1/1 fs/ocfs2: fix race in ocfs2_dentry_attach_lock
Hi Joseph,
Quoted 51 lines · Joseph Qi
Hi Wengang,
Quoted 48 lines · Wengang Wang
ocfs2_dentry_attach_lock()can be executed in parallel threads against the same dentry. Make that race safe.
The race is like this:thread A thread B
(A1) enter
ocfs2_dentry_attach_lock,
seeingdentry->d_fsdatais NULL,
and no alias found byocfs2_find_local_alias, so kmalloc
a newocfs2_dentry_lockstructure
to local variable "dl", dl1.....
(B1) enter
ocfs2_dentry_attach_lock, seeingdentry->d_fsdatais NULL, and no alias found byocfs2_find_local_aliasso kmalloc a newocfs2_dentry_lockstructure to local variable "dl", dl2.......
(A2) set
dentry->d_fsdatawith dl1,
callocfs2_dentry_lock()and increasedl1->dl_lockres.l_ro_holdersto 1 on
success....... (B2) set dentry->d_fsdata with dl2 call ocfs2_dentry_lock() and increase dl2->dl_lockres.l_ro_holders to 1 on success. ......(A3) call
ocfs2_dentry_unlock()
and decreasedl2->dl_lockres.l_ro_holdersto 0
on success..... (B3) call ocfs2_dentry_unlock(), decreasing dl2->dl_lockres.l_ro_holders, but see it's zero now, panic
Quoted 8 lines
Quoted 4 lines
Signed-off-by Wengang Wang <[email protected]>Reported-by Daniel Sobe <[email protected]>Tested-by Daniel Sobe <[email protected]>Reviewed-by Changwei Ge <[email protected]>Thanks for the detailed description.
IIUC, you are trying to identify the race case and free the second dentry lock.
Yes.
Quoted 8 lines
So would the following be more clear? Something like:
if (unlikely(dentry->d_fsdata && !alias)) { dl_attached = true; } else ... }
Quoted 8 lines
...
if (unlikely(dl_attached)) { ocfs2_lock_res_free(&dl->dl_lockres); kfree(dl); iput(inode); return 0; }
Seems your idea is to rename the variable "dl_free_on_race" to "dl_attached".
I think "dl_free_on_race" is more meaningful than "dl_attached" :D "dl_free_on_race" includes the following meanings:
0) is a dentry lock
1) need to be freed
2) on race condition only
By this name, we know it's a dentry lock that needs to be freed when race condition hits.
I don't exactly understand what "dl_attached" stands for, but from it's name, I am guessing the following:
0) it's dentry lock
1) it's attached to something
I don't think the dentry lock is attached anywhere yet, it's a local variable, no other reference so far. And why we want to free an "attached" object? Can you shall your thought for "dl_attached"?
thanks,
wengang
Quoted 51 lines
Thanks,
JosephQuoted 48 lines
---
v3: add Reviewed-by, Reported-by and Tested-by onlyv2: 1) removed lock on
dentry_attach_lockat the first access of
dentry->d_fsdatasince it helps very little.
2) do cleanups before freeing the duplicated dl
3) return after freeing the duplicated dl found.--- fs/ocfs2/dcache.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)fs/ocfs2/dcache.c +14
-2@@ -230,6 +230,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,int ret;struct dentry *alias;struct ocfs2_dentry_lock *dl = dentry->d_fsdata; +struct ocfs2_dentry_lock *dl_free_on_race =NULL ;trace_ocfs2_dentry_attach_lock (dentry->d_name.len, dentry->d_name.name, (unsigned long long )parent_blkno, dl);@@ -310,10 +311,21 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, out_attach:spin_lock (&dentry_attach_lock); - dentry->d_fsdata = dl; - dl->dl_count++; +/* d_fsdata could be set by parallel thread */ +if (unlikely (dentry->d_fsdata && !alias)) { + dl_free_on_race = dl; + }else { + dentry->d_fsdata = dl; + dl->dl_count++; + }spin_unlock (&dentry_attach_lock); +if (unlikely (dl_free_on_race)) { +iput (dl_free_on_race->dl_inode); +ocfs2_lock_res_free (&dl_free_on_race->dl_lockres); +kfree (dl_free_on_race); +return 0 ; + }/* * This actually gets us our PRMODE level lock. From now on, * we'll have a notification if one of these names is
PATCH v3 1/1 fs/ocfs2: fix race in ocfs2_dentry_attach_lock
Quoted 54 lines · Wengang Wang
Hi Joseph,
Quoted 51 lines · Joseph Qi
Hi Wengang,
Quoted 48 lines · Wengang Wang
ocfs2_dentry_attach_lock()can be executed in parallel threads against the same dentry. Make that race safe.
The race is like this:thread A thread B
(A1) enter
ocfs2_dentry_attach_lock,
seeingdentry->d_fsdatais NULL,
and no alias found byocfs2_find_local_alias, so kmalloc
a newocfs2_dentry_lockstructure
to local variable "dl", dl1.....
(B1) enter
ocfs2_dentry_attach_lock, seeingdentry->d_fsdatais NULL, and no alias found byocfs2_find_local_aliasso kmalloc a newocfs2_dentry_lockstructure to local variable "dl", dl2.......
(A2) set
dentry->d_fsdatawith dl1,
callocfs2_dentry_lock()and increasedl1->dl_lockres.l_ro_holdersto 1 on
success....... (B2) set dentry->d_fsdata with dl2 call ocfs2_dentry_lock() and increase dl2->dl_lockres.l_ro_holders to 1 on success. ......(A3) call
ocfs2_dentry_unlock()
and decreasedl2->dl_lockres.l_ro_holdersto 0
on success..... (B3) call ocfs2_dentry_unlock(), decreasing dl2->dl_lockres.l_ro_holders, but see it's zero now, panic
Quoted 12 lines
Quoted 7 lines
Quoted 4 lines
Signed-off-by Wengang Wang <[email protected]>Reported-by Daniel Sobe <[email protected]>Tested-by Daniel Sobe <[email protected]>Reviewed-by Changwei Ge <[email protected]>Thanks for the detailed description.
IIUC, you are trying to identify the race case and free the second dentry lock.Yes.
Quoted 4 lines
So would the following be more clear? Something like:
if (unlikely(dentry->d_fsdata && !alias)) { dl_attached = true;
Oh, seems "dl_attached" means d_fsdata is set with a dentry lock.
Well, this name may can't reflect the racing.? It can't tell it's set with new allocated dentry lock or with the one from alias dentry either, but it's set true when it's set with a new lock. Also, at the beginning of the function, we are checking if d_fsdata is set, in the d_fsdata set case, we don't set dl_attached, so I get the feeling that that variable is inconsistent.
But I don't care much about the this.? If you feel this way is better, choose a better name to replace dl_free_on_race.
thanks,
wengang
Quoted 88 lines
Quoted 4 lines
} else
... }Quoted 8 lines
...
if (unlikely(dl_attached)) { ocfs2_lock_res_free(&dl->dl_lockres); kfree(dl); iput(inode); return 0; }Seems your idea is to rename the variable "
dl_free_on_race" to "dl_attached".I think "
dl_free_on_race" is more meaningful than "dl_attached" :D "dl_free_on_race" includes the following meanings:
0) is a dentry lock
1) need to be freed
2) on race condition onlyBy this name, we know it's a dentry lock that needs to be freed when race condition hits.
I don't exactly understand what "
dl_attached" stands for, but from it's name, I am guessing the following:
0) it's dentry lock
1) it's attached to somethingI don't think the dentry lock is attached anywhere yet, it's a local variable, no other reference so far. And why we want to free an "attached" object? Can you shall your thought for "
dl_attached"?thanks,
wengangQuoted 41 lines
Thanks,
JosephQuoted 38 lines
---
v3: add Reviewed-by, Reported-by and Tested-by onlyv2: 1) removed lock on
dentry_attach_lockat the first access of
dentry->d_fsdatasince it helps very little.
2) do cleanups before freeing the duplicated dl
3) return after freeing the duplicated dl found.--- fs/ocfs2/dcache.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)fs/ocfs2/dcache.c +8
-2@@ -230,6 +230,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,int ret;struct dentry *alias;struct ocfs2_dentry_lock *dl = dentry->d_fsdata; +struct ocfs2_dentry_lock *dl_free_on_race =NULL ;trace_ocfs2_dentry_attach_lock (dentry->d_name.len, dentry->d_name.name, (unsigned long long )parent_blkno, dl);@@ -310,10 +311,21 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, out_attach:spin_lock (&dentry_attach_lock); - dentry->d_fsdata = dl; - dl->dl_count++; +/* d_fsdata could be set by parallel thread */ +if (unlikely (dentry->d_fsdata && !alias)) { + dl_free_on_race = dl; + }else { + dentry->d_fsdata = dl; + dl->dl_count++; + }spin_unlock (&dentry_attach_lock);Quoted 10 lines
fs/ocfs2/dcache.c int ocfs2_dentry_attach_lock(struct dentry *dentry, · line 331
Quoted 10 lines
fs/ocfs2/dcache.c +6
-0+if (unlikely (dl_free_on_race)) { +iput (dl_free_on_race->dl_inode); +ocfs2_lock_res_free (&dl_free_on_race->dl_lockres); +kfree (dl_free_on_race); +return 0 ; + }/* * This actually gets us our PRMODE level lock. From now on, * we'll have a notification if one of these names is _______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel
PATCH v3 1/1 fs/ocfs2: fix race in ocfs2_dentry_attach_lock
Quoted 63 lines · Wengang Wang
Quoted 61 lines · Wengang Wang
Hi Joseph,
Quoted 58 lines · Joseph Qi
Hi Wengang,
Quoted 48 lines · Wengang Wang
ocfs2_dentry_attach_lock()can be executed in parallel threads against the same dentry. Make that race safe.
The race is like this:????????????? thread A?????????????????????????????? thread B
(A1) enter
ocfs2_dentry_attach_lock,
seeingdentry->d_fsdatais NULL,
and no alias found byocfs2_find_local_alias, so kmalloc
a newocfs2_dentry_lockstructure
to local variable "dl", dl1???????????????? .....
????????????????????????????????????? (B1) enter
ocfs2_dentry_attach_lock, ????????????????????????????????????? seeingdentry->d_fsdatais NULL, ????????????????????????????????????? and no alias found by ?????????????????????????????????????ocfs2_find_local_aliasso kmalloc ????????????????????????????????????? a newocfs2_dentry_lockstructure ????????????????????????????????????? to local variable "dl", dl2.???????????????????????????????????????????????????? ......
(A2) set
dentry->d_fsdatawith dl1,
callocfs2_dentry_lock()and increasedl1->dl_lockres.l_ro_holdersto 1 on
success.
??????????????? ......????????????????????????????????????? (B2) set
dentry->d_fsdatawith dl2 ????????????????????????????????????? callocfs2_dentry_lock()and increase ???????????????????dl2->dl_lockres.l_ro_holdersto 1 on ??????????????????? success.??????????????????????????????????????????????????? ......
(A3) call
ocfs2_dentry_unlock()
and decreasedl2->dl_lockres.l_ro_holdersto 0
on success.
?????????????? ....????????????????????????????????????? (B3) call
ocfs2_dentry_unlock(), ????????????????????????????????????? decreasing ???????????????????dl2->dl_lockres.l_ro_holders, but ??????????????????? see it's zero now, panicQuoted 4 lines
Signed-off-by Wengang Wang <[email protected]>Reported-by Daniel Sobe <[email protected]>Tested-by Daniel Sobe <[email protected]>Reviewed-by Changwei Ge <[email protected]>Thanks for the detailed description.
IIUC, you are trying to identify the race case and free the second dentry lock.
Quoted 7 lines
Quoted 5 lines
Yes.
Quoted 4 lines
So would the following be more clear? Something like:
if (
unlikely(dentry->d_fsdata && !alias)) {
????dl_attached = true;Oh, seems "
dl_attached" meansd_fsdatais set with a dentry lock.
Yes, it means we've already attached the dentry lock. We can add some comments before setting it to true for code readability.
Quoted 4 lines
Well, this name may can't reflect the racing.? It can't tell it's set with new allocated dentry lock or with the one from alias dentry either, but it's set true when it's set with a new lock. Also, at the beginning of the function, we are checking if
d_fsdatais set, in thed_fsdataset case, we don't setdl_attached, so I get the feeling that that variable is inconsistent.But I don't care much about the this.? If you feel this way is better, choose a better name to replace
dl_free_on_race.
I don't meant 'dl_free_on_race' is not good.
I just don't want to introduce another dentry lock here since it's in fact the local dl itself.
Thanks,
Joseph
Quoted 90 lines
thanks,
wengangQuoted 85 lines
Quoted 4 lines
} else
????...
}Quoted 8 lines
...
if (
unlikely(dl_attached)) {
????ocfs2_lock_res_free(&dl->dl_lockres);
????kfree(dl);
????iput(inode);
????return 0;
}Seems your idea is to rename the variable "
dl_free_on_race" to "dl_attached".I think "
dl_free_on_race" is more meaningful than "dl_attached" :D "dl_free_on_race" includes the following meanings:
0) is a dentry lock
1) need to be freed
2) on race condition onlyBy this name, we know it's a dentry lock that needs to be freed when race condition hits.
I don't exactly understand what "
dl_attached" stands for, but from it's name, I am guessing the following:
0) it's dentry lock
1) it's attached to somethingI don't think the dentry lock is attached anywhere yet, it's a local variable, no other reference so far. And why we want to free an "attached" object? Can you shall your thought for "
dl_attached"?thanks,
wengangQuoted 40 lines
Thanks,
JosephQuoted 37 lines
---
v3: add Reviewed-by, Reported-by and Tested-by onlyv2: 1) removed lock on
dentry_attach_lockat the first access of ????????dentry->d_fsdatasince it helps very little. ????? 2) do cleanups before freeing the duplicated dl ????? 3) return after freeing the duplicated dl found.
---
??fs/ocfs2/dcache.c| 16 ++++++++++++++--
?? 1 file changed, 14 insertions(+), 2 deletions(-)fs/ocfs2/dcache.c +1
-0@@ -230,6 +230,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, ?????? int ret; ?????? struct dentry *alias; ?????? struct ocfs2_dentry_lock *dl = dentry->d_fsdata; +??? struct ocfs2_dentry_lock *dl_free_on_race = NULL; ?? ?????? trace_ocfs2_dentry_attach_lock(dentry->d_name.len, dentry->d_name.name, ????????????????????????? (unsigned long long)parent_blkno, dl);@@ -310,10 +311,21 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,?? ??
out_attach:
??????spin_lock(&dentry_attach_lock);
-???dentry->d_fsdata= dl;
-??? dl->dl_count++;
+??? /*d_fsdatacould be set by parallel thread */ +??? if (unlikely(dentry->d_fsdata && !alias)) { +???????dl_free_on_race= dl;
+??? } else {
+???????dentry->d_fsdata= dl;
+??????? dl->dl_count++;
+??? }
??????spin_unlock(&dentry_attach_lock);
?? +??? if (unlikely(dl_free_on_race)) { +???????iput(dl_free_on_race->dl_inode);Quoted 8 lines
Quoted 8 lines
+???????
ocfs2_lock_res_free(&dl_free_on_race->dl_lockres); +???????kfree(dl_free_on_race);
+??????? return 0;
+??? }
?????? /*
??????? * This actually gets us our PRMODE level lock. From now on, ??????? * we'll have a notification if one of these names is_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel
PATCH v3 1/1 fs/ocfs2: fix race in ocfs2_dentry_attach_lock
Quoted 74 lines · Joseph Qi
Quoted 62 lines · Wengang Wang
Quoted 61 lines · Wengang Wang
Hi Joseph,
Quoted 58 lines · Joseph Qi
Hi Wengang,
Quoted 48 lines · Wengang Wang
ocfs2_dentry_attach_lock()can be executed in parallel threads against the same dentry. Make that race safe.
The race is like this:????????????? thread A?????????????????????????????? thread B
(A1) enter
ocfs2_dentry_attach_lock,
seeingdentry->d_fsdatais NULL,
and no alias found byocfs2_find_local_alias, so kmalloc
a newocfs2_dentry_lockstructure
to local variable "dl", dl1???????????????? .....
????????????????????????????????????? (B1) enter
ocfs2_dentry_attach_lock, ????????????????????????????????????? seeingdentry->d_fsdatais NULL, ????????????????????????????????????? and no alias found by ?????????????????????????????????????ocfs2_find_local_aliasso kmalloc ????????????????????????????????????? a newocfs2_dentry_lockstructure ????????????????????????????????????? to local variable "dl", dl2.???????????????????????????????????????????????????? ......
(A2) set
dentry->d_fsdatawith dl1,
callocfs2_dentry_lock()and increasedl1->dl_lockres.l_ro_holdersto 1 on
success.
??????????????? ......????????????????????????????????????? (B2) set
dentry->d_fsdatawith dl2 ????????????????????????????????????? callocfs2_dentry_lock()and increase ???????????????????dl2->dl_lockres.l_ro_holdersto 1 on ??????????????????? success.??????????????????????????????????????????????????? ......
(A3) call
ocfs2_dentry_unlock()
and decreasedl2->dl_lockres.l_ro_holdersto 0
on success.
?????????????? ....????????????????????????????????????? (B3) call
ocfs2_dentry_unlock(), ????????????????????????????????????? decreasing ???????????????????dl2->dl_lockres.l_ro_holders, but ??????????????????? see it's zero now, panicQuoted 4 lines
Signed-off-by Wengang Wang <[email protected]>Reported-by Daniel Sobe <[email protected]>Tested-by Daniel Sobe <[email protected]>Reviewed-by Changwei Ge <[email protected]>Thanks for the detailed description.
IIUC, you are trying to identify the race case and free the second dentry lock.Quoted 6 lines
Quoted 5 lines
Yes.
Quoted 4 lines
So would the following be more clear? Something like:
if (
unlikely(dentry->d_fsdata && !alias)) {
????dl_attached = true;Oh, seems "
dl_attached" meansd_fsdatais set with a dentry lock.Yes, it means we've already attached the dentry lock. We can add some comments before setting it to true for code readability.
Quoted 6 lines
Quoted 3 lines
Well, this name may can't reflect the racing.? It can't tell it's set with new allocated dentry lock or with the one from alias dentry either, but it's set true when it's set with a new lock. Also, at the beginning of the function, we are checking if
d_fsdatais set, in thed_fsdataset case, we don't setdl_attached, so I get the feeling that that variable is inconsistent.But I don't care much about the this.? If you feel this way is better, choose a better name to replace
dl_free_on_race.I don't meant '
dl_free_on_race' is not good.
I just don't want to introduce another dentry lock here since it's in fact the local dl itself.
OK. I'd like another vote for this. If other guys also think introducing a bool variable is better than a dentry lock, I am going to revise it as v4; Otherwise, I'd like to keep it as v3.
thanks,
wengang
Quoted 92 lines
Thanks,
JosephQuoted 89 lines
thanks,
wengangQuoted 85 lines
Quoted 4 lines
} else
????...
}Quoted 8 lines
...
if (
unlikely(dl_attached)) {
????ocfs2_lock_res_free(&dl->dl_lockres);
????kfree(dl);
????iput(inode);
????return 0;
}Seems your idea is to rename the variable "
dl_free_on_race" to "dl_attached".I think "
dl_free_on_race" is more meaningful than "dl_attached" :D "dl_free_on_race" includes the following meanings:
0) is a dentry lock
1) need to be freed
2) on race condition onlyBy this name, we know it's a dentry lock that needs to be freed when race condition hits.
I don't exactly understand what "
dl_attached" stands for, but from it's name, I am guessing the following:
0) it's dentry lock
1) it's attached to somethingI don't think the dentry lock is attached anywhere yet, it's a local variable, no other reference so far. And why we want to free an "attached" object? Can you shall your thought for "
dl_attached"?thanks,
wengangQuoted 40 lines
Thanks,
JosephQuoted 37 lines
---
v3: add Reviewed-by, Reported-by and Tested-by onlyv2: 1) removed lock on
dentry_attach_lockat the first access of
????????dentry->d_fsdatasince it helps very little. ????? 2) do cleanups before freeing the duplicated dl ????? 3) return after freeing the duplicated dl found.--- ?? fs/ocfs2/dcache.c | 16 ++++++++++++++--?? 1 file changed, 14 insertions(+), 2 deletions(-)
fs/ocfs2/dcache.c +9
-2@@ -230,6 +230,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, ?????? int ret; ?????? struct dentry *alias; ?????? struct ocfs2_dentry_lock *dl = dentry->d_fsdata; +??? struct ocfs2_dentry_lock *dl_free_on_race = NULL; ?? ?????? trace_ocfs2_dentry_attach_lock(dentry->d_name.len, dentry->d_name.name, ????????????????????????? (unsigned long long)parent_blkno, dl);@@ -310,10 +311,21 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, ?? ?? out_attach: ?????? spin_lock(&dentry_attach_lock); -??? dentry->d_fsdata = dl; -??? dl->dl_count++; +??? /* d_fsdata could be set by parallel thread */ +??? if (unlikely(dentry->d_fsdata && !alias)) { +??????? dl_free_on_race = dl; +??? } else { +??????? dentry->d_fsdata = dl; +??????? dl->dl_count++; +??? } ?????? spin_unlock(&dentry_attach_lock); ?? +??? if (unlikely(dl_free_on_race)) { +??????? iput(dl_free_on_race->dl_inode);Quoted 8 lines
fs/ocfs2/dcache.c int ocfs2_dentry_attach_lock(struct dentry *dentry, · line 330
Quoted 8 lines
fs/ocfs2/dcache.c +4
-0+??????? ocfs2_lock_res_free(&dl_free_on_race->dl_lockres); +??????? kfree(dl_free_on_race); +??????? return 0; +??? } ?????? /*??????? * This actually gets us our PRMODE level lock. From now on, ??????? * we'll have a notification if one of these names is _______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel