[PATCH 1/2] nfs/nfsd: return err directly while malloc failing

[PATCH 1/2] nfs/nfsd: return err directly while malloc failing

[Rui Xiang]

while nthreads malloc fails, return err value directory
without kfree.

And the same issue arises in nfs_lookup, so return res
directly without freeing fh and fattr.

Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
---
 fs/nfs/dir.c     | 4 ++--
 fs/nfsd/nfsctl.c | 7 ++++---
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 812154a..2518865 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1278,7 +1278,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
 
 	res = ERR_PTR(-ENAMETOOLONG);
 	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
-		goto out;
+		return res;
 
 	/*
 	 * If we're doing an exclusive create, optimize away the lookup
@@ -1287,7 +1287,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
 	if (nfs_is_exclusive_create(dir, flags)) {
 		d_instantiate(dentry, NULL);
 		res = NULL;
-		goto out;
+		return res;
 	}
 
 	res = ERR_PTR(-ENOMEM);
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index 7f55517..3c5e748 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -491,9 +491,10 @@ static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
 	}
 
 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
-	rv = -ENOMEM;
-	if (nthreads == NULL)
-		goto out_free;
+	if (nthreads == NULL) {
+		mutex_unlock(&nfsd_mutex);
+		return -ENOENT;
+	}
 
 	if (size > 0) {
 		for (i = 0; i < npools; i++) {
-- 
1.8.2.2

[PATCH 2/2] nfs: fix return err if inode exiting in nfs_instantiate

[Rui Xiang]

In common function nfs_instantiate to create, mkdir, and mknod,
if dentry->d_inode exits, it should return -EEXIST instead of
-EACCES.

Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
---
 fs/nfs/dir.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 2518865..e570b37 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1547,7 +1547,7 @@ int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
 	struct dentry *parent = dget_parent(dentry);
 	struct inode *dir = parent->d_inode;
 	struct inode *inode;
-	int error = -EACCES;
+	int error = -EEXIST;
 
 	d_drop(dentry);
 
-- 
1.8.2.2

Re: [PATCH 2/2] nfs: fix return err if inode exiting in nfs_instantiate

[Trond Myklebust]

On Dec 6, 2013, at 4:30, Rui Xiang <rui.xiang@huawei.com> wrote:

> In common function nfs_instantiate to create, mkdir, and mknod,
> if dentry->d_inode exits, it should return -EEXIST instead of
> -EACCES.
> 
> Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
> ---
> fs/nfs/dir.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index 2518865..e570b37 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -1547,7 +1547,7 @@ int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
> 	struct dentry *parent = dget_parent(dentry);
> 	struct inode *dir = parent->d_inode;
> 	struct inode *inode;
> -	int error = -EACCES;
> +	int error = -EEXIST;
> 
> 	d_drop(sentry);
> 

That looks like it should rather be a WARN_ON(). If the caller has set the dentry's inode before creating the file, then something is really wrong.

Cheers
  Trond

Re: [PATCH 2/2] nfs: fix return err if inode exiting in nfs_instantiate

[Rui Xiang]

On 2013/12/6 21:46, Trond Myklebust wrote:
> 
> On Dec 6, 2013, at 4:30, Rui Xiang <rui.xiang@huawei.com> wrote:
> 
>> In common function nfs_instantiate to create, mkdir, and mknod,
>> if dentry->d_inode exits, it should return -EEXIST instead of
>> -EACCES.
>>
>> Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
>> ---
>> fs/nfs/dir.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
>> index 2518865..e570b37 100644
>> --- a/fs/nfs/dir.c
>> +++ b/fs/nfs/dir.c
>> @@ -1547,7 +1547,7 @@ int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
>> 	struct dentry *parent = dget_parent(dentry);
>> 	struct inode *dir = parent->d_inode;
>> 	struct inode *inode;
>> -	int error = -EACCES;
>> +	int error = -EEXIST;
>>
>> 	d_drop(dentry);
>>
> 
> That looks like it should rather be a WARN_ON(). If the caller has set the dentry's inode before creating the file, then something is really wrong.
It can return -EEXIST to say that the file already exits, then will exit. But why does it need a WARN_ON(). Please give me some advise.

Thanks
 Rui
> 
> Cheers
>   Trond
> 

Re: [PATCH 2/2] nfs: fix return err if inode exiting in nfs_instantiate

[Trond Myklebust]

On Mon, 2013-12-09 at 18:42 +0800, Rui Xiang wrote:
> On 2013/12/6 21:46, Trond Myklebust wrote:
> > 
> > On Dec 6, 2013, at 4:30, Rui Xiang <rui.xiang@huawei.com> wrote:
> > 
> >> In common function nfs_instantiate to create, mkdir, and mknod,
> >> if dentry->d_inode exits, it should return -EEXIST instead of
> >> -EACCES.
> >>
> >> Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
> >> ---
> >> fs/nfs/dir.c | 2 +-
> >> 1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> >> index 2518865..e570b37 100644
> >> --- a/fs/nfs/dir.c
> >> +++ b/fs/nfs/dir.c
> >> @@ -1547,7 +1547,7 @@ int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
> >> 	struct dentry *parent = dget_parent(dentry);
> >> 	struct inode *dir = parent->d_inode;
> >> 	struct inode *inode;
> >> -	int error = -EACCES;
> >> +	int error = -EEXIST;
> >>
> >> 	d_drop(dentry);
> >>
> > 
> > That looks like it should rather be a WARN_ON(). If the caller has set the dentry's inode before creating the file, then something is really wrong.
> It can return -EEXIST to say that the file already exits, then will exit. But why does it need a WARN_ON(). Please give me some advise.
> 

If we ever hit that condition then it means that the caller is doing
something very wrong. That's what the WARN_ON (or WARN_ON_ONCE) would be
there to check.

Cheers
  Trond

Re: [PATCH 2/2] nfs: fix return err if inode exiting in nfs_instantiate

[Rui Xiang]

On 2013/12/11 0:26, Trond Myklebust wrote:
> On Mon, 2013-12-09 at 18:42 +0800, Rui Xiang wrote:
>> On 2013/12/6 21:46, Trond Myklebust wrote:
>>>
>>> On Dec 6, 2013, at 4:30, Rui Xiang <rui.xiang@huawei.com> wrote:
>>>
>>>> In common function nfs_instantiate to create, mkdir, and mknod,
>>>> if dentry->d_inode exits, it should return -EEXIST instead of
>>>> -EACCES.
>>>>
>>>> Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
>>>> ---
>>>> fs/nfs/dir.c | 2 +-
>>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
>>>> index 2518865..e570b37 100644
>>>> --- a/fs/nfs/dir.c
>>>> +++ b/fs/nfs/dir.c
>>>> @@ -1547,7 +1547,7 @@ int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
>>>> 	struct dentry *parent = dget_parent(dentry);
>>>> 	struct inode *dir = parent->d_inode;
>>>> 	struct inode *inode;
>>>> -	int error = -EACCES;
>>>> +	int error = -EEXIST;
>>>>
>>>> 	d_drop(dentry);
>>>>
>>>
>>> That looks like it should rather be a WARN_ON(). If the caller has set the dentry's inode before creating the file, then something is really wrong.
>> It can return -EEXIST to say that the file already exits, then will exit. But why does it need a WARN_ON(). Please give me some advise.
>>
> 
> If we ever hit that condition then it means that the caller is doing
> something very wrong. That's what the WARN_ON (or WARN_ON_ONCE) would be
> there to check.
> 
OK. In that case, is it necessary to send a new patch with the WARN_ON?


Thanks
 Rui

Re: [PATCH 1/2] nfs/nfsd: return err directly while malloc failing

[J. Bruce Fields]

On Fri, Dec 06, 2013 at 05:30:16PM +0800, Rui Xiang wrote:
> while nthreads malloc fails, return err value directory
> without kfree.
> 
> And the same issue arises in nfs_lookup, so return res
> directly without freeing fh and fattr.
> 
> Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
> ---
>  fs/nfs/dir.c     | 4 ++--
>  fs/nfsd/nfsctl.c | 7 ++++---
>  2 files changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index 812154a..2518865 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -1278,7 +1278,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
>  
>  	res = ERR_PTR(-ENAMETOOLONG);
>  	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
> -		goto out;
> +		return res;
>  
>  	/*
>  	 * If we're doing an exclusive create, optimize away the lookup
> @@ -1287,7 +1287,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
>  	if (nfs_is_exclusive_create(dir, flags)) {
>  		d_instantiate(dentry, NULL);
>  		res = NULL;
> -		goto out;
> +		return res;
>  	}
>  
>  	res = ERR_PTR(-ENOMEM);
> diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> index 7f55517..3c5e748 100644
> --- a/fs/nfsd/nfsctl.c
> +++ b/fs/nfsd/nfsctl.c
> @@ -491,9 +491,10 @@ static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
>  	}
>  
>  	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
> -	rv = -ENOMEM;
> -	if (nthreads == NULL)
> -		goto out_free;
> +	if (nthreads == NULL) {
> +		mutex_unlock(&nfsd_mutex);
> +		return -ENOENT;
> +	}
>  
>  	if (size > 0) {
>  		for (i = 0; i < npools; i++) {
> -- 

Looks fine.  Could you send this last chunk to me as a separate patch?

--b.

> 1.8.2.2
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/2] nfs/nfsd: return err directly while malloc failing

[Rui Xiang]

On 2013/12/6 17:30, Rui Xiang wrote:
> while nthreads malloc fails, return err value directory
> without kfree.
> 
> And the same issue arises in nfs_lookup, so return res
> directly without freeing fh and fattr.
> 
> Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
> ---
>  fs/nfs/dir.c     | 4 ++--
>  fs/nfsd/nfsctl.c | 7 ++++---
>  2 files changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index 812154a..2518865 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -1278,7 +1278,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
>  
>  	res = ERR_PTR(-ENAMETOOLONG);
>  	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
> -		goto out;
> +		return res;
>  
>  	/*
>  	 * If we're doing an exclusive create, optimize away the lookup
> @@ -1287,7 +1287,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
>  	if (nfs_is_exclusive_create(dir, flags)) {
>  		d_instantiate(dentry, NULL);
>  		res = NULL;
> -		goto out;
> +		return res;
>  	}
>  
>  	res = ERR_PTR(-ENOMEM);
Trond,  and what's your opinion on the chunk above?

Thanks
 Rui

> diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> index 7f55517..3c5e748 100644
> --- a/fs/nfsd/nfsctl.c
> +++ b/fs/nfsd/nfsctl.c
> @@ -491,9 +491,10 @@ static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
>  	}
>  
>  	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
> -	rv = -ENOMEM;
> -	if (nthreads == NULL)
> -		goto out_free;
> +	if (nthreads == NULL) {
> +		mutex_unlock(&nfsd_mutex);
> +		return -ENOENT;
> +	}
>  
>  	if (size > 0) {
>  		for (i = 0; i < npools; i++) {
> 

Re: [PATCH 1/2] nfs/nfsd: return err directly while malloc failing

[Trond Myklebust]

On Mon, 2013-12-09 at 20:33 +0800, Rui Xiang wrote:
> On 2013/12/6 17:30, Rui Xiang wrote:
> > while nthreads malloc fails, return err value directory
> > without kfree.
> > 
> > And the same issue arises in nfs_lookup, so return res
> > directly without freeing fh and fattr.
> > 
> > Signed-off-by: Rui Xiang <rui.xiang@huawei.com>
> > ---
> >  fs/nfs/dir.c     | 4 ++--
> >  fs/nfsd/nfsctl.c | 7 ++++---
> >  2 files changed, 6 insertions(+), 5 deletions(-)
> > 
> > diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> > index 812154a..2518865 100644
> > --- a/fs/nfs/dir.c
> > +++ b/fs/nfs/dir.c
> > @@ -1278,7 +1278,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
> >  
> >  	res = ERR_PTR(-ENAMETOOLONG);
> >  	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
> > -		goto out;
> > +		return res;
> >  
> >  	/*
> >  	 * If we're doing an exclusive create, optimize away the lookup
> > @@ -1287,7 +1287,7 @@ struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned in
> >  	if (nfs_is_exclusive_create(dir, flags)) {
> >  		d_instantiate(dentry, NULL);
> >  		res = NULL;
> > -		goto out;
> > +		return res;
> >  	}
> >  
> >  	res = ERR_PTR(-ENOMEM);
> Trond,  and what's your opinion on the chunk above?

It's a minor optimisation but certainly not something that needs to be
pushed in before the next merge window.

Cheers
  Trond