Variable `now` is declared without initialization. The variable
could be accessed inside the if-else statements following the
variable declaration, before it has been initialized.

This patch initializes the variable to
`inode_set_ctime_current(inode)` by default.

This issue was reported by Coverity Scan.

Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
---
 fs/attr.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/fs/attr.c b/fs/attr.c
index c614b954bda5..77523af2e62d 100644
--- a/fs/attr.c
+++ b/fs/attr.c
@@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
 static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
 {
 	unsigned int ia_valid = attr->ia_valid;
-	struct timespec64 now;
+	struct timespec64 now = inode_set_ctime_current(inode);
 
 	if (ia_valid & ATTR_CTIME) {
 		/*
@@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
 		 */
 		if (ia_valid & ATTR_DELEG)
 			now = inode_set_ctime_deleg(inode, attr->ia_ctime);
-		else
-			now = inode_set_ctime_current(inode);
 	} else {
 		/* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
 		WARN_ON_ONCE(ia_valid & ATTR_MTIME);
-- 
2.43.0

On Wed, Oct 09, 2024 at 02:05:25PM -0600, Everest K.C. wrote:
> Variable `now` is declared without initialization. The variable
> could be accessed inside the if-else statements following the
> variable declaration, before it has been initialized.
> 
> This patch initializes the variable to
> `inode_set_ctime_current(inode)` by default.
> 
> This issue was reported by Coverity Scan.
> 
> Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>

Fixes: d8d11298e8a1 ("fs: handle delegated timestamps in setattr_copy_mgtime")

Maybe the WARN_ON_ONCE() should be updated to check ATTR_ATIME as well?

regards,
dan carpenter

> ---
>  fs/attr.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/fs/attr.c b/fs/attr.c
> index c614b954bda5..77523af2e62d 100644
> --- a/fs/attr.c
> +++ b/fs/attr.c
> @@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
>  static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
>  {
>  	unsigned int ia_valid = attr->ia_valid;
> -	struct timespec64 now;
> +	struct timespec64 now = inode_set_ctime_current(inode);
>  
>  	if (ia_valid & ATTR_CTIME) {
>  		/*
> @@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
>  		 */
>  		if (ia_valid & ATTR_DELEG)
>  			now = inode_set_ctime_deleg(inode, attr->ia_ctime);
> -		else
> -			now = inode_set_ctime_current(inode);
>  	} else {
>  		/* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
>  		WARN_ON_ONCE(ia_valid & ATTR_MTIME);
> -- 
> 2.43.0
>

On Wed, 2024-10-09 at 23:45 +0300, Dan Carpenter wrote:
> On Wed, Oct 09, 2024 at 02:05:25PM -0600, Everest K.C. wrote:
> > Variable `now` is declared without initialization. The variable
> > could be accessed inside the if-else statements following the
> > variable declaration, before it has been initialized.
> > 
> > This patch initializes the variable to
> > `inode_set_ctime_current(inode)` by default.
> > 
> > This issue was reported by Coverity Scan.
> > 
> > Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
> 
> Fixes: d8d11298e8a1 ("fs: handle delegated timestamps in setattr_copy_mgtime")
> 
> Maybe the WARN_ON_ONCE() should be updated to check ATTR_ATIME as well?
> 
> regards,
> dan carpenter
> 
> > ---
> >  fs/attr.c | 4 +---
> >  1 file changed, 1 insertion(+), 3 deletions(-)
> > 
> > diff --git a/fs/attr.c b/fs/attr.c
> > index c614b954bda5..77523af2e62d 100644
> > --- a/fs/attr.c
> > +++ b/fs/attr.c
> > @@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
> >  static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> >  {
> >  	unsigned int ia_valid = attr->ia_valid;
> > -	struct timespec64 now;
> > +	struct timespec64 now = inode_set_ctime_current(inode);
> >  
> >  	if (ia_valid & ATTR_CTIME) {
> >  		/*
> > @@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> >  		 */
> >  		if (ia_valid & ATTR_DELEG)
> >  			now = inode_set_ctime_deleg(inode, attr->ia_ctime);
> > -		else
> > -			now = inode_set_ctime_current(inode);
> >  	} else {
> >  		/* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
> >  		WARN_ON_ONCE(ia_valid & ATTR_MTIME);
> > -- 
> > 2.43.0
> > 

This doesn't look correct. inode_set_ctime_current will update the time
to the current time and then inode_set_ctime_deleg() won't work
properly if the time being set is earlier than that.

I proposed a different fix earlier today which should be correct:

https://lore.kernel.org/linux-fsdevel/20241009-mgtime-v1-1-383b9e0481b5@kernel.org/

Thanks,
-- 
Jeff Layton <jlayton@kernel.org>

On Wed, Oct 9, 2024 at 4:40 PM Jeff Layton <jlayton@kernel.org> wrote:
>
> On Wed, 2024-10-09 at 23:45 +0300, Dan Carpenter wrote:
> > On Wed, Oct 09, 2024 at 02:05:25PM -0600, Everest K.C. wrote:
> > > Variable `now` is declared without initialization. The variable
> > > could be accessed inside the if-else statements following the
> > > variable declaration, before it has been initialized.
> > >
> > > This patch initializes the variable to
> > > `inode_set_ctime_current(inode)` by default.
> > >
> > > This issue was reported by Coverity Scan.
> > >
> > > Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
> >
> > Fixes: d8d11298e8a1 ("fs: handle delegated timestamps in setattr_copy_mgtime")
> >
> > Maybe the WARN_ON_ONCE() should be updated to check ATTR_ATIME as well?
> >
> > regards,
> > dan carpenter
> >
> > > ---
> > >  fs/attr.c | 4 +---
> > >  1 file changed, 1 insertion(+), 3 deletions(-)
> > >
> > > diff --git a/fs/attr.c b/fs/attr.c
> > > index c614b954bda5..77523af2e62d 100644
> > > --- a/fs/attr.c
> > > +++ b/fs/attr.c
> > > @@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
> > >  static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> > >  {
> > >     unsigned int ia_valid = attr->ia_valid;
> > > -   struct timespec64 now;
> > > +   struct timespec64 now = inode_set_ctime_current(inode);
> > >
> > >     if (ia_valid & ATTR_CTIME) {
> > >             /*
> > > @@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> > >              */
> > >             if (ia_valid & ATTR_DELEG)
> > >                     now = inode_set_ctime_deleg(inode, attr->ia_ctime);
> > > -           else
> > > -                   now = inode_set_ctime_current(inode);
> > >     } else {
> > >             /* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
> > >             WARN_ON_ONCE(ia_valid & ATTR_MTIME);
> > > --
> > > 2.43.0
> > >
>
> This doesn't look correct. inode_set_ctime_current will update the time
> to the current time and then inode_set_ctime_deleg() won't work
> properly if the time being set is earlier than that.
Yes. I get it now.
> I proposed a different fix earlier today which should be correct:
>
> https://lore.kernel.org/linux-fsdevel/20241009-mgtime-v1-1-383b9e0481b5@kernel.org/
> Thanks,
> --
> Jeff Layton <jlayton@kernel.org>
Thanks,
Everest K.C.

On Wed, Oct 9, 2024 at 2:45 PM Dan Carpenter <dan.carpenter@linaro.org> wrote:
>
> On Wed, Oct 09, 2024 at 02:05:25PM -0600, Everest K.C. wrote:
> > Variable `now` is declared without initialization. The variable
> > could be accessed inside the if-else statements following the
> > variable declaration, before it has been initialized.
> >
> > This patch initializes the variable to
> > `inode_set_ctime_current(inode)` by default.
> >
> > This issue was reported by Coverity Scan.
> >
> > Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
>
> Fixes: d8d11298e8a1 ("fs: handle delegated timestamps in setattr_copy_mgtime")
>
> Maybe the WARN_ON_ONCE() should be updated to check ATTR_ATIME as well?
I am not sure about that, but even if that is necessary. I think it
should be handled in a different patch.
> regards,
> dan carpenter
>
> > ---
> >  fs/attr.c | 4 +---
> >  1 file changed, 1 insertion(+), 3 deletions(-)
> >
> > diff --git a/fs/attr.c b/fs/attr.c
> > index c614b954bda5..77523af2e62d 100644
> > --- a/fs/attr.c
> > +++ b/fs/attr.c
> > @@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
> >  static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> >  {
> >       unsigned int ia_valid = attr->ia_valid;
> > -     struct timespec64 now;
> > +     struct timespec64 now = inode_set_ctime_current(inode);
> >
> >       if (ia_valid & ATTR_CTIME) {
> >               /*
> > @@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> >                */
> >               if (ia_valid & ATTR_DELEG)
> >                       now = inode_set_ctime_deleg(inode, attr->ia_ctime);
> > -             else
> > -                     now = inode_set_ctime_current(inode);
> >       } else {
> >               /* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
> >               WARN_ON_ONCE(ia_valid & ATTR_MTIME);
> > --
> > 2.43.0
> >

On 10/9/24 14:05, Everest K.C. wrote:
> Variable `now` is declared without initialization. The variable
> could be accessed inside the if-else statements following the
> variable declaration, before it has been initialized.

It could be, but it isn't. I am not sure if this change is needed.
> 
> This patch initializes the variable to
> `inode_set_ctime_current(inode)` by default.

Instead of "This patch initializes", change it to "Initialize ..."
Do refer to submitting patches document for information on how
to write change logs.

> 
> This issue was reported by Coverity Scan.

Include the the error/report from Coverity.

> 
> Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
> ---
>   fs/attr.c | 4 +---
>   1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/fs/attr.c b/fs/attr.c
> index c614b954bda5..77523af2e62d 100644
> --- a/fs/attr.c
> +++ b/fs/attr.c
> @@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
>   static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
>   {
>   	unsigned int ia_valid = attr->ia_valid;
> -	struct timespec64 now;
> +	struct timespec64 now = inode_set_ctime_current(inode);
>   
>   	if (ia_valid & ATTR_CTIME) {
>   		/*
> @@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
>   		 */
>   		if (ia_valid & ATTR_DELEG)
>   			now = inode_set_ctime_deleg(inode, attr->ia_ctime);
> -		else
> -			now = inode_set_ctime_current(inode);

The code is clear and easy to read the way it is since it handles both cases
and does appropriate initialization.


>   	} else {
>   		/* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
>   		WARN_ON_ONCE(ia_valid & ATTR_MTIME);

I will leave it up to the maintainers to decide whether to take
this change or not.

thanks,
-- Shuah

On Wed, Oct 9, 2024 at 2:38 PM Shuah Khan <skhan@linuxfoundation.org> wrote:
>
> On 10/9/24 14:05, Everest K.C. wrote:
> > Variable `now` is declared without initialization. The variable
> > could be accessed inside the if-else statements following the
> > variable declaration, before it has been initialized.
>
> It could be, but it isn't. I am not sure if this change is needed.
If you look at the full code then,  if  `ia_valid & ATTR_CTIME`
evaluates to False then now is never initialized.

> > This patch initializes the variable to
> > `inode_set_ctime_current(inode)` by default.
>
> Instead of "This patch initializes", change it to "Initialize ..."
> Do refer to submitting patches document for information on how
> to write change logs.
Will do that and send V2.
> >
> > This issue was reported by Coverity Scan.
>
> Include the the error/report from Coverity.
Will do that and send V2.
> >
> > Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
> > ---
> >   fs/attr.c | 4 +---
> >   1 file changed, 1 insertion(+), 3 deletions(-)
> >
> > diff --git a/fs/attr.c b/fs/attr.c
> > index c614b954bda5..77523af2e62d 100644
> > --- a/fs/attr.c
> > +++ b/fs/attr.c
> > @@ -284,7 +284,7 @@ EXPORT_SYMBOL(inode_newsize_ok);
> >   static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> >   {
> >       unsigned int ia_valid = attr->ia_valid;
> > -     struct timespec64 now;
> > +     struct timespec64 now = inode_set_ctime_current(inode);
> >
> >       if (ia_valid & ATTR_CTIME) {
> >               /*
> > @@ -293,8 +293,6 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
> >                */
> >               if (ia_valid & ATTR_DELEG)
> >                       now = inode_set_ctime_deleg(inode, attr->ia_ctime);
> > -             else
> > -                     now = inode_set_ctime_current(inode);
>
> The code is clear and easy to read the way it is since it handles both cases
> and does appropriate initialization.
Yes, I agree, but if we initialize now to the current time during its
declaration then the else
condition won't be necessary.
>
> >       } else {
> >               /* If ATTR_CTIME isn't set, then ATTR_MTIME shouldn't be either. */
> >               WARN_ON_ONCE(ia_valid & ATTR_MTIME);
>
> I will leave it up to the maintainers to decide whether to take
> this change or not.
>
> thanks,
> -- Shuah