On (14/12/15 13:34), Lukas Slebodnik wrote:
> On (14/12/15 12:38), Pavel Březina wrote:
>> On 12/14/2015 12:05 PM, Lukas Slebodnik wrote:
>>> On (11/12/15 15:56), Pavel Březina wrote:
>>>> On 12/11/2015 03:07 PM, Jakub Hrozek wrote:
>>>>>> From 2479352011f96a8f64c403d1278b791c73df703a Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Fri, 6 Nov 2015 13:00:47 +0100
>>>>>> Subject: [PATCH 01/35] SUDO: convert periodical refreshes to
be_ptask
>>>>>
>>>>> LGTM, not tested yet. One question inline...
>>>>>
>>>>>>
>>>>>> This removes old sudo timer and simplyfies code a lot. It also
>>>>>> allows to manage offline/online state.
>>>>>>
>>>>>> - Full and smart refresh are disabled when offline.
>>>>>> - Full refresh is run immediately when sssd is back online.
>>>>>> - Smart refresh is scheduled normally when sssd is back online.
>>>>>>
>>>>>> Resolves:
>>>>>>
https://fedorahosted.org/sssd/ticket/1943
>>>>>
>>>>> I agree with the logic, but it's different from the original
code, so is
>>>>> the change intended?
>>>>
>>>> Yes, the functionality remains the same but we handle online/offline
>>>> transition much better here.
>>>>
>>>>>> From d17b4d1c63f1b85a14a6fe8b2400e2c29158cd43 Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Fri, 6 Nov 2015 13:32:34 +0100
>>>>>> Subject: [PATCH 02/35] SUDO: move refreshes from sdap_sudo.c to
>>>>>> sdap_sudo_refresh.c
>>>>>
>>>>> LGTM, I wonder why git doesn't show the changes as move since
such a
>>>>> substantial part of sdap_sudo.c changed?
>>>>
>>>> Dunno.
>>>>
>>>>>> From 633d0de6e3edab4d5b988506426800cf35845702 Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Mon, 9 Nov 2015 12:26:07 +0100
>>>>>> Subject: [PATCH 03/35] SUDO: move offline check to handler
>>>>>>
>>>>>> We let sdap_id_op decide if we are offline or not here but we
>>>>>> should not get to this code since ptask is disabled and we will
>>>>>> not get through sudo handler if offline.
>>>>>
>>>>> LGTM, the connection error would still take us offline, which is what
we
>>>>> want.
>>>>>
>>>>>> From 3dc1bbb6fa4353f0bb2cc2b9417900a6d4695240 Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Mon, 9 Nov 2015 12:44:13 +0100
>>>>>> Subject: [PATCH 04/35] SUDO: simplify error handling
>>>>>
>>>>> LGTM
>>>>>
>>>>>> From 2e85798a9d1f856b549cb733b6a7f353bec05e6e Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Mon, 9 Nov 2015 12:52:35 +0100
>>>>>> Subject: [PATCH 05/35] SUDO: fix sdap_id_op logic
>>>>>
>>>>> LGTM, the fail/immediate confusion in sdap_sudo_load_sudoers_send is
>>>>> fixed in another patch. If you end up changing this patchset, it
might
>>>>> also be better to fix the "i" iterators to use size_t
instead of int,
>>>>> but if you don't change them, it's fine.
>>>>
>>>> Ok, new patch is attached to the patch set. Thanks.
>>>>
>>>>>
>>>>>> From 03bcc4d6b5c1324644bab7b98a10bfacaf9096c7 Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Mon, 9 Nov 2015 13:34:45 +0100
>>>>>> Subject: [PATCH 07/35] SUDO: fix sdap_sudo_smart_refresh_recv()
>>>>>>
>>>>>> This fix huge violation of tevent coding style.
>>>>>
>>>>> LGTM
>>>>>
>>>>>> From 0093a630eac9a745512c643b1389506c3839a56e Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Mon, 9 Nov 2015 13:59:48 +0100
>>>>>> Subject: [PATCH 08/35] SUDO: sdap_sudo_load_sudoers improve
iterator
>>>>>>
>>>>>> The old search base iterator was difficult to read since its
logic
>>>>>> spread through all functions. This patch also shorten names.
>>>>>
>>>>> LGTM
>>>>>
>>>>>> From a7d3e06ac32b74e2586d87f9671b5a49eda7cafe Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Tue, 10 Nov 2015 10:34:41 +0100
>>>>>> Subject: [PATCH 09/35] SUDO: set USN inside sdap_sudo_refresh
request
>>>>>>
>>>>>> Reduce code duplication.
>>>>>> ---
>>>>>> src/providers/ldap/sdap_async_sudo.c | 48
+++++++++++++++++++++------
>>>>>> src/providers/ldap/sdap_sudo.h | 2 +-
>>>>>> src/providers/ldap/sdap_sudo_refresh.c | 59
++++++----------------------------
>>>>>> 3 files changed, 49 insertions(+), 60 deletions(-)
>>>>>>
>>>>>> diff --git a/src/providers/ldap/sdap_async_sudo.c
b/src/providers/ldap/sdap_async_sudo.c
>>>>>> index
71e1b3322a557beaf82b51316967b3147fb61f76..7a90ccde9cbd2efc0d0509f21e072cd745c57844 100644
>>>>>> --- a/src/providers/ldap/sdap_async_sudo.c
>>>>>> +++ b/src/providers/ldap/sdap_async_sudo.c
>>>>>> @@ -283,6 +283,7 @@ static int sdap_sudo_store_sudoers(TALLOC_CTX
*mem_ctx,
>>>>>>
>>>>>> /* Empty sudoers? Done. */
>>>>>> if (rules_count == 0 || rules == NULL) {
>>>>>> + *_usn = NULL;
>>>>>> return EOK;
>>>>>> }
>>>>>>
>>>>>> @@ -299,8 +300,37 @@ static int
sdap_sudo_store_sudoers(TALLOC_CTX *mem_ctx,
>>>>>> return EOK;
>>>>>> }
>>>>>>
>>>>>> +static void sdap_sudo_set_usn(struct sdap_server_opts *srv_opts,
char *usn)
>>>>>> +{
>>>>>> + unsigned int usn_number;
>>>>>> + char *endptr = NULL;
>>>>>> +
>>>>>> + if (usn == NULL) {
>>>>>> + DEBUG(SSSDBG_TRACE_FUNC, "Empty USN,
ignoring\n");
>>>>>> + return;
>>>>>> + }
>>>>>> +
>>>>>> + if (srv_opts == NULL) {
>>>>>> + DEBUG(SSSDBG_TRACE_FUNC, "Bug: srv_opts is
NULL\n");
>>>>>> + return;
>>>>>> + }
>>>>>> +
>>>>>> + talloc_zfree(srv_opts->max_sudo_value);
>>>>>> + srv_opts->max_sudo_value = talloc_steal(srv_opts, usn);
>>>>>> +
>>>>>> + usn_number = strtoul(usn, &endptr, 10);
>>>>>> + if ((endptr == NULL || (*endptr == '\0' &&
endptr != usn))
>>>>>> + && (usn_number > srv_opts->last_usn)) {
>>>>>> + srv_opts->last_usn = usn_number;
>>>>>> + }
>>>>>> +
>>>>>> + DEBUG(SSSDBG_FUNC_DATA, "SUDO higher USN value:
[%s]\n",
>>>>>> + srv_opts->max_sudo_value);
>>>>>> +}
>>>>>> +
>>>>>
>>>>> What is the logic behind this function? Why do we store the USN both
as
>>>>> a string unconditionally and as a number conditionally? Wouldn't
it make
>>>>> sense to just store the data in one format?
>>>>
>>>> max_sudo_value is internal to sudo, it is max value found in sudo
subtree.
>>>> It's string since it is not needed to be a number. last_usn is the
largers
>>>> value on the whole server, it needs to be compared.
>>>>
>>>> max_sudo_value can be a number also, but it would be inconsistent with
other
>>>> providers.
>>>>
>>>>>
>>>>>> From 1ca17bd4110960f3d9b6834446cec83dfe9a1ca8 Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Tue, 10 Nov 2015 11:31:50 +0100
>>>>>> Subject: [PATCH 10/35] SUDO: built host filter inside
sdap_sudo_refresh
>>>>>> request
>>>>>>
>>>>>> Preparation for:
>>>>>>
https://fedorahosted.org/sssd/ticket/2672
>>>>>
>>>>> LGTM
>>>>>
>>>>>> From ebe26c922ad72c97f53d5ca0f5eeef1034c4a470 Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Tue, 10 Nov 2015 11:34:14 +0100
>>>>>> Subject: [PATCH 11/35] SUDO: fail if usn is unknown in smart
refresh
>>>>>>
>>>>>> USN should never be unknown when we run smart refresh so this is
>>>>>> more a safety condition for a code bug. Nevertheless, we should
>>>>>> not immitate full refresh here under those conditions.
>>>>>
>>>>> LGTM
>>>>>
>>>>>> From aa06528d2a1fdd6b5df5de629d2da5297cbcc90d Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Tue, 10 Nov 2015 12:34:11 +0100
>>>>>> Subject: [PATCH 12/35] SUDO: fix potential memory leak in
sdap_sudo_init
>>>>>
>>>>> LGTM
>>>>>
>>>>>> From 897f68a3379f8fab70eae106f5fa71b4fe4d809a Mon Sep 17
00:00:00 2001
>>>>>> From: =?UTF-8?q?Pavel=20B=C5=99ezina?=
<pbrezina(a)redhat.com>
>>>>>> Date: Tue, 10 Nov 2015 12:32:16 +0100
>>>>>> Subject: [PATCH 13/35] SUDO: obtain host information when going
online
>>>>>>
>>>>>> Resolves:
>>>>>>
https://fedorahosted.org/sssd/ticket/2672
>>>>>
>>>>> [...]
>>>>>
>>>>>> +static void sdap_sudo_refresh_hostinfo_done(struct tevent_req
*subreq)
>>>>>> +{
>>>>>> + struct sdap_sudo_ctx *sudo_ctx;
>>>>>> + struct sdap_sudo_refresh_state *state;
>>>>>> + struct tevent_req *req;
>>>>>> + errno_t ret;
>>>>>> +
>>>>>> + req = tevent_req_callback_data(subreq, struct tevent_req);
>>>>>> + state = tevent_req_data(req, struct
sdap_sudo_refresh_state);
>>>>>> +
>>>>>> + sudo_ctx = state->sudo_ctx;
>>>>>> +
>>>>>> + ret = sdap_sudo_get_hostinfo_recv(sudo_ctx, subreq,
&sudo_ctx->hostnames,
>>>>>> +
&sudo_ctx->ip_addr);
>>>>>> + talloc_zfree(subreq);
>>>>>> + if (ret != EOK) {
>>>>>> + DEBUG(SSSDBG_OP_FAILURE, "Unable to retrieve host
information, "
>>>>>> + "host filter will be
disabled [%d]: %s\n",
>>>>>> + ret, sss_strerror(ret));
>>>>>> + sudo_ctx->use_host_filter = false;
>>>>>> + }
>>>>>
>>>>> I think you forgot an else here..
>>>>>
>>>>
>>>> Yes. thanks.
>>>>
>>>
>>> We had a discussion on IRC with Pavel and Jakub
>>> and patch set caused a bug. The smart refresh was not executed
>>> if USN was NULL. (full refresh did not found any sudo rules)
>>>
>>> e.g.
>>> * install freeIPA (sssd is started)
>>> * add sudo rules
>>> * sudo rules will be downloaded after full refresh
>>> (which is 1 day by default or after restarting sssd)
>>>
>>> LS
>>
>> I have changed the patch 0009 in a way so we do not imitate full refresh in
>> smart refresh but still be able to run smart refresh if no rules were found.
>>
> Here is a diff:
> diff --git a/src/providers/ldap/sdap_async_sudo.c
b/src/providers/ldap/sdap_async_sudo.c
> index 9f66d6c..13824c5 100644
> --- a/src/providers/ldap/sdap_async_sudo.c
> +++ b/src/providers/ldap/sdap_async_sudo.c
> @@ -305,13 +305,21 @@ static void sdap_sudo_set_usn(struct sdap_server_opts
*srv_opts, char *usn)
> unsigned int usn_number;
> char *endptr = NULL;
>
> - if (usn == NULL) {
> - DEBUG(SSSDBG_TRACE_FUNC, "Empty USN, ignoring\n");
> + if (srv_opts == NULL) {
> + DEBUG(SSSDBG_TRACE_FUNC, "Bug: srv_opts is NULL\n");
> return;
> }
>
> - if (srv_opts == NULL) {
> - DEBUG(SSSDBG_TRACE_FUNC, "Bug: srv_opts is NULL\n");
> + if (usn == NULL) {
> + /* If the USN value is unknown and we don't have max_sudo_value set
> + * (possibly first full refresh which did not find any rule) we will
> + * set zero so smart refresh can pick up. */
> + if (srv_opts->max_sudo_value == NULL) {
> + srv_opts->max_sudo_value = "0";
I'm sorry that I didn't noticed it earlier.
You assign static allocated constant here.
But there is zfree :-) It was not visible from diff
313 if (usn == NULL) {
314 /* If the USN value is unknown and we don't have max_sudo_value set
315 * (possibly first full refresh which did not find any rule) we will
316 * set zero so smart refresh can pick up. */
317 if (srv_opts->max_sudo_value == NULL) {
318 srv_opts->max_sudo_value = "0";
319 return;
320 }
321
322 DEBUG(SSSDBG_TRACE_FUNC, "Empty USN, ignoring\n");
323 return;
324 }
325
326 talloc_zfree(srv_opts->max_sudo_value);
327 srv_opts->max_sudo_value = talloc_steal(srv_opts, usn);
==35602== Process terminating with default action of signal 11 (SIGSEGV)
==35602== General Protection Fault
==35602== at 0x3EB8447E2C: vfprintf (vfprintf.c:1641)
==35602== by 0x3EB84FFB0F: __vsnprintf_chk (vsnprintf_chk.c:65)
==35602== by 0x3EBE405402: talloc_vasprintf (stdio2.h:78)
==35602== by 0x3EBE402BDC: talloc_log (talloc.c:288)
==35602== by 0x3EBE402B13: _talloc_free (talloc.c:356)
==35602== by 0xD038168: sdap_sudo_refresh_done (sdap_async_sudo.c:326)
==35602== by 0xD037BCE: sdap_sudo_load_sudoers_done (sdap_async_sudo.c:195)
==35602== by 0xD005516: generic_ext_search_handler (sdap_async.c:1677)
==35602== by 0xD009AA2: sdap_get_generic_op_finished (sdap_async.c:1603)
==35602== by 0xD00AD0E: sdap_process_result (sdap_async.c:352)
==35602== by 0x3EC1007C90: tevent_common_loop_timer_delay (tevent_timed.c:341)
==35602== by 0x3EC1008CBA: epoll_event_loop_once (tevent_epoll.c:916)
--35602-- Discarding syms at 0xd8c41f0-0xd8cc648 in /lib64/libnss_files-2.12.so due to
munmap()
--35602-- Discarding syms at 0xdad1000-0xdad4388 in /lib64/libnss_dns-2.12.so due to
munmap()
--35602-- Discarding syms at 0xf06c760-0xf0717f8 in /lib64/libnss_sss.so.2 due to
munmap()
LS
Thanks. I should pay more attention when doing a quick fix, I was
already overcoded though. I hope this patch set is the one :-)