2022-08-26 10:43:44

by Christophe JAILLET

Subject: [PATCH] nfsd: Fix a memory leak in an error handling path

If this memdup_user() call fails, the memory allocated in a previous call
a few lines above should be freed. Otherwise it leaks.

Fixes: 6ee95d1c8991 ("nfsd: add support for upcall version 2")
Signed-off-by: Christophe JAILLET <[email protected]>
---
Speculative, untested.
---
fs/nfsd/nfs4recover.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c
index b29d27eaa8a6..248ff9f4141c 100644
--- a/fs/nfsd/nfs4recover.c
+++ b/fs/nfsd/nfs4recover.c
@@ -815,8 +815,10 @@ __cld_pipe_inprogress_downcall(const struct cld_msg_v2 __user *cmsg,
princhash.data = memdup_user(
&ci->cc_princhash.cp_data,
princhashlen);
- if (IS_ERR_OR_NULL(princhash.data))
+ if (IS_ERR_OR_NULL(princhash.data)) {
+ kfree(name.data);
return -EFAULT;
+ }
princhash.len = princhashlen;
} else
princhash.len = 0;
--
2.34.1

2022-08-26 10:58:00

by Jeffrey Layton

Subject: Re: [PATCH] nfsd: Fix a memory leak in an error handling path

On Fri, 2022-08-26 at 12:24 +0200, Christophe JAILLET wrote:
> If this memdup_user() call fails, the memory allocated in a previous call
> a few lines above should be freed. Otherwise it leaks.
>
> Fixes: 6ee95d1c8991 ("nfsd: add support for upcall version 2")
> Signed-off-by: Christophe JAILLET <[email protected]>
> ---
> Speculative, untested.
> ---
> fs/nfsd/nfs4recover.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c
> index b29d27eaa8a6..248ff9f4141c 100644
> --- a/fs/nfsd/nfs4recover.c
> +++ b/fs/nfsd/nfs4recover.c
> @@ -815,8 +815,10 @@ __cld_pipe_inprogress_downcall(const struct cld_msg_v2 __user *cmsg,
> princhash.data = memdup_user(
> &ci->cc_princhash.cp_data,
> princhashlen);
> - if (IS_ERR_OR_NULL(princhash.data))
> + if (IS_ERR_OR_NULL(princhash.data)) {
> + kfree(name.data);
> return -EFAULT;
> + }
> princhash.len = princhashlen;
> } else
> princhash.len = 0;

Reviewed-by: Jeff Layton <[email protected]>

2022-08-26 11:10:05

by Dan Carpenter

Subject: Re: [PATCH] nfsd: Fix a memory leak in an error handling path

On Fri, Aug 26, 2022 at 12:24:54PM +0200, Christophe JAILLET wrote:
> If this memdup_user() call fails, the memory allocated in a previous call
> a few lines above should be freed. Otherwise it leaks.
>
> Fixes: 6ee95d1c8991 ("nfsd: add support for upcall version 2")
> Signed-off-by: Christophe JAILLET <[email protected]>
> ---
> Speculative, untested.
> ---
> fs/nfsd/nfs4recover.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c
> index b29d27eaa8a6..248ff9f4141c 100644
> --- a/fs/nfsd/nfs4recover.c
> +++ b/fs/nfsd/nfs4recover.c
> @@ -815,8 +815,10 @@ __cld_pipe_inprogress_downcall(const struct cld_msg_v2 __user *cmsg,
> princhash.data = memdup_user(
> &ci->cc_princhash.cp_data,
> princhashlen);
> - if (IS_ERR_OR_NULL(princhash.data))
> + if (IS_ERR_OR_NULL(princhash.data)) {
> + kfree(name.data);
> return -EFAULT;

This comment is not directed at you and is not related to your patch.
But memdup_user() never returns NULL, only error pointers. I wrote a
fifteen page blog entry about NULL vs error pointers the other week.
https://staticthinking.wordpress.com/2022/08/01/mixing-error-pointers-and-null/
This should propagate the error code from memdup_user() instead of
-EFAULT.

regards,
dan carpenter

2022-08-30 21:15:55

by Chuck Lever

Subject: Re: [PATCH] nfsd: Fix a memory leak in an error handling path

> On Aug 26, 2022, at 7:08 AM, Dan Carpenter <[email protected]> wrote:
>
> On Fri, Aug 26, 2022 at 12:24:54PM +0200, Christophe JAILLET wrote:
>> If this memdup_user() call fails, the memory allocated in a previous call
>> a few lines above should be freed. Otherwise it leaks.
>>
>> Fixes: 6ee95d1c8991 ("nfsd: add support for upcall version 2")
>> Signed-off-by: Christophe JAILLET <[email protected]>
>> ---
>> Speculative, untested.
>> ---
>> fs/nfsd/nfs4recover.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c
>> index b29d27eaa8a6..248ff9f4141c 100644
>> --- a/fs/nfsd/nfs4recover.c
>> +++ b/fs/nfsd/nfs4recover.c
>> @@ -815,8 +815,10 @@ __cld_pipe_inprogress_downcall(const struct cld_msg_v2 __user *cmsg,
>> princhash.data = memdup_user(
>> &ci->cc_princhash.cp_data,
>> princhashlen);
>> - if (IS_ERR_OR_NULL(princhash.data))
>> + if (IS_ERR_OR_NULL(princhash.data)) {
>> + kfree(name.data);
>> return -EFAULT;
>
> This comment is not directed at you and is not related to your patch.
> But memdup_user() never returns NULL, only error pointers. I wrote a
> fifteen page blog entry about NULL vs error pointers the other week.
> https://staticthinking.wordpress.com/2022/08/01/mixing-error-pointers-and-null/
> This should propagate the error code from memdup_user() instead of
> -EFAULT.

I take it then that Christophe should redrive this with your suggested
corrections? I haven't applied this yet because I was waiting for
follow-up.


--
Chuck Lever

2022-08-31 05:50:58

by Dan Carpenter

Subject: Re: [PATCH] nfsd: Fix a memory leak in an error handling path

On Tue, Aug 30, 2022 at 09:11:54PM +0000, Chuck Lever III wrote:
>
>
> > On Aug 26, 2022, at 7:08 AM, Dan Carpenter <[email protected]> wrote:
> >
> > On Fri, Aug 26, 2022 at 12:24:54PM +0200, Christophe JAILLET wrote:
> >> If this memdup_user() call fails, the memory allocated in a previous call
> >> a few lines above should be freed. Otherwise it leaks.
> >>
> >> Fixes: 6ee95d1c8991 ("nfsd: add support for upcall version 2")
> >> Signed-off-by: Christophe JAILLET <[email protected]>
> >> ---
> >> Speculative, untested.
> >> ---
> >> fs/nfsd/nfs4recover.c | 4 +++-
> >> 1 file changed, 3 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c
> >> index b29d27eaa8a6..248ff9f4141c 100644
> >> --- a/fs/nfsd/nfs4recover.c
> >> +++ b/fs/nfsd/nfs4recover.c
> >> @@ -815,8 +815,10 @@ __cld_pipe_inprogress_downcall(const struct cld_msg_v2 __user *cmsg,
> >> princhash.data = memdup_user(
> >> &ci->cc_princhash.cp_data,
> >> princhashlen);
> >> - if (IS_ERR_OR_NULL(princhash.data))
> >> + if (IS_ERR_OR_NULL(princhash.data)) {
> >> + kfree(name.data);
> >> return -EFAULT;
> >
> > This comment is not directed at you and is not related to your patch.
> > But memdup_user() never returns NULL, only error pointers. I wrote a
> > fifteen page blog entry about NULL vs error pointers the other week.
> > https://staticthinking.wordpress.com/2022/08/01/mixing-error-pointers-and-null/
> > This should propagate the error code from memdup_user() instead of
> > -EFAULT.
>
> I take it then that Christophe should redrive this with your suggested
> corrections? I haven't applied this yet because I was waiting for
> follow-up.

No, that's a different thing. If anyone wants to clean that up then it
should be part of a different patch.

regards,
dan carpenter