On Fri, Aug 16, 2013 at 06:35:52PM +0200, Sumit Bose wrote:
> Hi,
>
> the second patch fixes
https://fedorahosted.org/sssd/ticket/1630 while
> the first adds a missing talloc_free(). One might argue that both
> talloc_free() can be removed because the parent context is freed later
> during the request.
>
> bye,
> Sumit
> From 8af12d60f892d25808a1a74517b9c2c2ff40eadf Mon Sep 17 00:00:00 2001
> From: Sumit Bose <sbose(a)redhat.com>
> Date: Tue, 13 Aug 2013 17:59:11 +0200
> Subject: [PATCH 1/2] ipa_s2n_get_user_done: free group_attrs as well
>
> ---
> src/providers/ipa/ipa_s2n_exop.c | 1 +
> 1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/src/providers/ipa/ipa_s2n_exop.c b/src/providers/ipa/ipa_s2n_exop.c
> index e16ec14..53746ca 100644
> --- a/src/providers/ipa/ipa_s2n_exop.c
> +++ b/src/providers/ipa/ipa_s2n_exop.c
> @@ -882,6 +882,7 @@ static void ipa_s2n_get_user_done(struct tevent_req *subreq)
>
> done:
> talloc_free(user_attrs);
> + talloc_free(group_attrs);
> if (ret == EOK) {
> tevent_req_done(req);
> } else {
> --
> 1.7.7.6
>
ACK. I think if the request is not expected to be a long-running one
(like a nested group request) we can get away with simply using state.
> From 7c54aef6677a8498907d57156f661992d91055e9 Mon Sep 17 00:00:00 2001
> From: Sumit Bose <sbose(a)redhat.com>
> Date: Tue, 13 Aug 2013 17:59:42 +0200
> Subject: [PATCH 2/2] ipa_s2n_get_user_done: make sure ALIAS name is lower
> case
>
> Fixes
https://fedorahosted.org/sssd/ticket/1630
> ---
> src/providers/ipa/ipa_s2n_exop.c | 20 ++++++++++++++++++--
> 1 files changed, 18 insertions(+), 2 deletions(-)
>
> diff --git a/src/providers/ipa/ipa_s2n_exop.c b/src/providers/ipa/ipa_s2n_exop.c
> index 53746ca..d8506aa 100644
> --- a/src/providers/ipa/ipa_s2n_exop.c
> +++ b/src/providers/ipa/ipa_s2n_exop.c
> @@ -651,6 +651,7 @@ static void ipa_s2n_get_user_done(struct tevent_req *subreq)
> struct sysdb_attrs *user_attrs = NULL;
> struct sysdb_attrs *group_attrs = NULL;
> char *name;
> + char *lc_name;
> char *realm;
> char *upn;
> struct berval *bv_req = NULL;
> @@ -766,7 +767,14 @@ static void ipa_s2n_get_user_done(struct tevent_req *subreq)
> goto done;
> }
>
> - ret = sysdb_attrs_add_string(user_attrs, SYSDB_NAME_ALIAS, name);
> + lc_name = sss_tc_utf8_str_tolower(user_attrs, name);
> + if (lc_name == NULL) {
> + DEBUG(SSSDBG_CRIT_FAILURE, ("Cannot convert name to
lowercase\n"));
> + ret = ENOMEM;
> + goto done;
> + }
> +
> + ret = sysdb_attrs_add_string(user_attrs, SYSDB_NAME_ALIAS, lc_name);
> if (ret != EOK) {
> DEBUG(SSSDBG_OP_FAILURE, ("sysdb_attrs_add_string
failed.\n"));
> goto done;
> @@ -844,8 +852,16 @@ static void ipa_s2n_get_user_done(struct tevent_req *subreq)
> goto done;
> }
>
> + lc_name = sss_tc_utf8_str_tolower(group_attrs, name);
> + if (lc_name == NULL) {
> + DEBUG(SSSDBG_CRIT_FAILURE,
> + ("Cannot convert name to lowercase\n"));
> + ret = ENOMEM;
> + goto done;
> + }
> +
> ret = sysdb_attrs_add_string(group_attrs, SYSDB_NAME_ALIAS,
> - name);
> + lc_name);
> if (ret != EOK) {
> DEBUG(SSSDBG_OP_FAILURE, ("sysdb_attrs_add_string
failed.\n"));
> goto done;
> --
> 1.7.7.6
>
ACK. I haven't been able to reproduce the issue because in my
configuration (F-19) the name I got from extop was already lowercased. I
think the bug depended on some version of Winbind? Anyway, the code
works fine, the nameAlias attribute is populated and is lowercased.