[PATCH] tools: Fix realloc() use in fdarray__grow()

[PATCH] tools: Fix realloc() use in fdarray__grow()

[Jann Horn]

If `entries != NULL`, then `fda->entries` has been freed, so whatever
happens afterwards, we must store `entries` in `fda->entries`.
If we bail out at the second realloc(), the new allocation will be bigger
than what fda->nr_alloc says, but that's fine.

Fixes: 2171a9256862 ("tools lib fd array: Allow associating an integer cookie with each entry")
Signed-off-by: Jann Horn <jannh@google.com>
---
To the maintainer:
I'm not sure about the etiquette for using CC stable in
patches for somewhat theoretical issues in userland tools;
feel free to tack a CC stable onto this if you think it
should go into stable.

 tools/lib/api/fd/array.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/tools/lib/api/fd/array.c b/tools/lib/api/fd/array.c
index 58d44d5eee31..acf8eca1a94a 100644
--- a/tools/lib/api/fd/array.c
+++ b/tools/lib/api/fd/array.c
@@ -27,15 +27,13 @@ int fdarray__grow(struct fdarray *fda, int nr)
 
 	if (entries == NULL)
 		return -ENOMEM;
+	fda->entries = entries;
 
 	priv = realloc(fda->priv, psize);
-	if (priv == NULL) {
-		free(entries);
+	if (priv == NULL)
 		return -ENOMEM;
-	}
 
 	fda->nr_alloc = nr_alloc;
-	fda->entries  = entries;
 	fda->priv     = priv;
 	return 0;
 }
-- 
2.25.0

Re: [PATCH] tools: Fix realloc() use in fdarray__grow()

[Jiri Olsa]

On Sat, Feb 29, 2020 at 05:26:07PM +0100, Jann Horn wrote:
> If `entries != NULL`, then `fda->entries` has been freed, so whatever
> happens afterwards, we must store `entries` in `fda->entries`.
> If we bail out at the second realloc(), the new allocation will be bigger
> than what fda->nr_alloc says, but that's fine.
> 
> Fixes: 2171a9256862 ("tools lib fd array: Allow associating an integer cookie with each entry")
> Signed-off-by: Jann Horn <jannh@google.com>
> ---
> To the maintainer:
> I'm not sure about the etiquette for using CC stable in
> patches for somewhat theoretical issues in userland tools;
> feel free to tack a CC stable onto this if you think it
> should go into stable.
> 
>  tools/lib/api/fd/array.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/tools/lib/api/fd/array.c b/tools/lib/api/fd/array.c
> index 58d44d5eee31..acf8eca1a94a 100644
> --- a/tools/lib/api/fd/array.c
> +++ b/tools/lib/api/fd/array.c
> @@ -27,15 +27,13 @@ int fdarray__grow(struct fdarray *fda, int nr)
>  
>  	if (entries == NULL)
>  		return -ENOMEM;
> +	fda->entries = entries;
>  
>  	priv = realloc(fda->priv, psize);
> -	if (priv == NULL) {
> -		free(entries);

so we are sure we always call fdarray__exit even
if we fail in here?  if that's the case then

Acked-by: Jiri Olsa <jolsa@redhat.com>

thanks,
jirka

> +	if (priv == NULL)
>  		return -ENOMEM;
> -	}
>  
>  	fda->nr_alloc = nr_alloc;
> -	fda->entries  = entries;
>  	fda->priv     = priv;
>  	return 0;
>  }
> -- 
> 2.25.0
> 

Re: [PATCH] tools: Fix realloc() use in fdarray__grow()

[Jann Horn]

On Mon, Mar 2, 2020 at 1:58 PM Jiri Olsa <jolsa@redhat.com> wrote:
> On Sat, Feb 29, 2020 at 05:26:07PM +0100, Jann Horn wrote:
> > If `entries != NULL`, then `fda->entries` has been freed, so whatever
> > happens afterwards, we must store `entries` in `fda->entries`.
> > If we bail out at the second realloc(), the new allocation will be bigger
> > than what fda->nr_alloc says, but that's fine.
> >
> > Fixes: 2171a9256862 ("tools lib fd array: Allow associating an integer cookie with each entry")
> > Signed-off-by: Jann Horn <jannh@google.com>
> > ---
> > To the maintainer:
> > I'm not sure about the etiquette for using CC stable in
> > patches for somewhat theoretical issues in userland tools;
> > feel free to tack a CC stable onto this if you think it
> > should go into stable.
> >
> >  tools/lib/api/fd/array.c | 6 ++----
> >  1 file changed, 2 insertions(+), 4 deletions(-)
> >
> > diff --git a/tools/lib/api/fd/array.c b/tools/lib/api/fd/array.c
> > index 58d44d5eee31..acf8eca1a94a 100644
> > --- a/tools/lib/api/fd/array.c
> > +++ b/tools/lib/api/fd/array.c
> > @@ -27,15 +27,13 @@ int fdarray__grow(struct fdarray *fda, int nr)
> >
> >       if (entries == NULL)
> >               return -ENOMEM;
> > +     fda->entries = entries;
> >
> >       priv = realloc(fda->priv, psize);
> > -     if (priv == NULL) {
> > -             free(entries);
>
> so we are sure we always call fdarray__exit even
> if we fail in here?  if that's the case then
>
> Acked-by: Jiri Olsa <jolsa@redhat.com>

Ugh... actually, no, at least fdarray__new() does a plain free().
While other places like FDA_ADD use fdarray_delete()...