Re: getpwnam(NULL) get "segmentation fault"

[Andrew Pinski <pinskia at gmail dot com>]

On Sat, May 30, 2015 at 3:52 PM, Sheng Yong <shengyong1@huawei.com> wrote:
> Hi,
>
> When using getpwnam(), we found that if the parameter is NULL, the program
> would get a segmentation fault. I checked the glibc code, and found that
> parameter passed to function is defined as a macro ADD_PARAMS and its value
> is not checked before using.

>
> It seems serval functions under nss directory doing so. But segmentation
> fault is not friendly to users. Maybe we could add some parameter-check
> code before using theses parameters like the following. But I don't know
> if it is reasonable to do that. Any hint is appreciated.



I don't see anywhere in POSIX standard where it says this is well
defined.  We have many different places were we don't check for NULL
pointers including and not limited to memcpy, strcmp, strcpy.  So in
my mind, this should fall in the same idea.

Thanks,
Andrew Pinski

>
> thanks,
> Sheng
>
>
> diff --git a/nss/getXXbyYY.c b/nss/getXXbyYY.c
> index 15fecf8..4ea7ee2 100644
> --- a/nss/getXXbyYY.c
> +++ b/nss/getXXbyYY.c
> @@ -96,6 +96,8 @@ FUNCTION_NAME (ADD_PARAMS)
>    /* Get lock.  */
>    __libc_lock_lock (lock);
>
> +  CHECK_PARAMS();
> +
>    if (buffer == NULL)
>      {
>        buffer_size = BUFLEN;
> diff --git a/pwd/getpwnam.c b/pwd/getpwnam.c
> index 9ec66d7..cb28279 100644
> --- a/pwd/getpwnam.c
> +++ b/pwd/getpwnam.c
> @@ -25,5 +25,9 @@
>  #define ADD_PARAMS     const char *name
>  #define ADD_VARIABLES  name
>  #define BUFLEN         NSS_BUFLEN_PASSWD
> +#define CHECK_PARAMS() do {    \
> +  if (name == NULL)            \
> +    return NULL;               \
> +} while (0)
>
>  #include "../nss/getXXbyYY.c"
>