On Tue, Feb 12, 2019 at 4:20 PM Petr Lautrbach <plautrba@xxxxxxxxxx> wrote:
>
> getgrnam_r() uses a preallocated buffer to store a structure containing
> the broken-out fields of the record in the group database. The size of
> this buffer is usually sysconf(_SC_GETGR_R_SIZE_MAX) == 1024 and it is
> not enough for groups with a large number of users. In these cases,
> getgrnam_r() returns -1 and sets errno to ERANGE and the caller can
> retry with a larger buffer.
>
> Fixes:
> $ semanage login -a -s user_u -r s0-s0:c1.c2 '%largegroup'
> libsemanage.semanage_direct_commit: semanage_genhomedircon returned error code -1. (Numerical result out of range).
> OSError: Numerical result out of range
>
> Signed-off-by: Petr Lautrbach <plautrba@xxxxxxxxxx>
Acked-by: Nicolas Iooss <nicolas.iooss@xxxxxxx>
> ---
> libsemanage/src/genhomedircon.c | 20 +++++++++++++++++---
> 1 file changed, 17 insertions(+), 3 deletions(-)
>
> diff --git a/libsemanage/src/genhomedircon.c b/libsemanage/src/genhomedircon.c
> index 591941fb..e5f8d371 100644
> --- a/libsemanage/src/genhomedircon.c
> +++ b/libsemanage/src/genhomedircon.c
> @@ -1077,10 +1077,24 @@ static int get_group_users(genhomedircon_settings_t * s,
>
> const char *grname = selogin + 1;
>
> - if (getgrnam_r(grname, &grstorage, grbuf,
> - (size_t) grbuflen, &group) != 0) {
> - goto cleanup;
> + errno = 0;
> + while (
> + (retval = getgrnam_r(grname, &grstorage, grbuf, (size_t) grbuflen, &group)) != 0 &&
> + errno == ERANGE
> + ) {
> + char *new_grbuf;
> + grbuflen *= 2;
> + if (grbuflen < 0)
> + /* the member list could exceed 2Gb on a system with a 32-bit CPU (where
> + * sizeof(long) = 4) - if this ever happened, the loop would become infinite. */
> + goto cleanup;
> + new_grbuf = realloc(grbuf, grbuflen);
> + if (new_grbuf == NULL)
> + goto cleanup;
> + grbuf = new_grbuf;
> }
> + if (retval != 0)
> + goto cleanup;
>
> if (group == NULL) {
> ERR(s->h_semanage, "Can't find group named %s\n", grname);
> --
> 2.20.1
>
