rpms/389-ds-base
6
0
Fork 0
Code Issues Pull Requests Releases Activity

Import from CS git

Browse Source
This commit is contained in:
eabdullin 2025-02-04 07:17:12 +00:00
parent 27a2aba684
commit fdd80910b8
3 changed files with 333 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

297
SOURCES/0021-Issue-6509-Race-condition-with-Paged-Result-searches.patch Normal file
View File

@ -0,0 +1,297 @@
From d2f9dd82e3610ee9b73feea981c680c03bb21394 Mon Sep 17 00:00:00 2001
From: Mark Reynolds <mreynolds@redhat.com>
Date: Thu, 16 Jan 2025 08:42:53 -0500
Subject: [PATCH] Issue 6509 - Race condition with Paged Result searches
Description:
There is a race condition with Paged Result searches when a new operation comes
in while a paged search is finishing. This triggers an invalid time out error
and closes the connection with a T3 code.
The problem is that we do not use the "PagedResult lock" when checking the
connection's paged result data for a timeout event. This causes the paged
result timeout value to change unexpectedly and trigger a false timeout when a
new operation arrives.
Now we check the timeout without hte conn lock, if its expired it could
be a race condition and false positive. Try the lock again and test the
timeout. This also prevents blocking non-paged result searches from
getting held up by the lock when it's not necessary.
This also fixes some memory leaks that occur when an error happens.
Relates: https://github.com/389ds/389-ds-base/issues/6509
Reviewed by: tbordaz & proger (Thanks!!)
---
ldap/servers/slapd/daemon.c | 61 ++++++++++++++++++-------------
ldap/servers/slapd/opshared.c | 58 ++++++++++++++---------------
ldap/servers/slapd/pagedresults.c | 9 +++++
ldap/servers/slapd/slap.h | 2 +-
4 files changed, 75 insertions(+), 55 deletions(-)
diff --git a/ldap/servers/slapd/daemon.c b/ldap/servers/slapd/daemon.c
index bb80dae36..13dfe250d 100644
--- a/ldap/servers/slapd/daemon.c
+++ b/ldap/servers/slapd/daemon.c
@@ -1578,7 +1578,29 @@ setup_pr_read_pds(Connection_Table *ct)
if (c->c_state == CONN_STATE_FREE) {
connection_table_move_connection_out_of_active_list(ct, c);
} else {
- /* we try to acquire the connection mutex, if it is already
+ /* Check for a timeout for PAGED RESULTS */
+ if (pagedresults_is_timedout_nolock(c)) {
+ /*
+ * There could be a race condition so lets try again with the
+ * right lock
+ */
+ pthread_mutex_t *pr_mutex = pageresult_lock_get_addr(c);
+ if (pthread_mutex_trylock(pr_mutex) == EBUSY) {
+ c = next;
+ continue;
+ }
+ if (pagedresults_is_timedout_nolock(c)) {
+ pthread_mutex_unlock(pr_mutex);
+ disconnect_server(c, c->c_connid, -1,
+ SLAPD_DISCONNECT_PAGED_SEARCH_LIMIT,
+ 0);
+ } else {
+ pthread_mutex_unlock(pr_mutex);
+ }
+ }
+
+ /*
+ * we try to acquire the connection mutex, if it is already
* acquired by another thread, don't wait
*/
if (pthread_mutex_trylock(&(c->c_mutex)) == EBUSY) {
@@ -1586,35 +1608,24 @@ setup_pr_read_pds(Connection_Table *ct)
continue;
}
if (c->c_flags & CONN_FLAG_CLOSING) {
- /* A worker thread has marked that this connection
- * should be closed by calling disconnect_server.
- * move this connection out of the active list
- * the last thread to use the connection will close it
+ /*
+ * A worker thread, or paged result timeout, has marked that
+ * this connection should be closed by calling
+ * disconnect_server(). Move this connection out of the active
+ * list then the last thread to use the connection will close
+ * it.
*/
connection_table_move_connection_out_of_active_list(ct, c);
} else if (c->c_sd == SLAPD_INVALID_SOCKET) {
connection_table_move_connection_out_of_active_list(ct, c);
} else if (c->c_prfd != NULL) {
if ((!c->c_gettingber) && (c->c_threadnumber < c->c_max_threads_per_conn)) {
- int add_fd = 1;
- /* check timeout for PAGED RESULTS */
- if (pagedresults_is_timedout_nolock(c)) {
- /* Exceeded the paged search timelimit; disconnect the client */
- disconnect_server_nomutex(c, c->c_connid, -1,
- SLAPD_DISCONNECT_PAGED_SEARCH_LIMIT,
- 0);
- connection_table_move_connection_out_of_active_list(ct,
- c);
- add_fd = 0; /* do not poll on this fd */
- }
- if (add_fd) {
- ct->fd[count].fd = c->c_prfd;
- ct->fd[count].in_flags = SLAPD_POLL_FLAGS;
- /* slot i of the connection table is mapped to slot
- * count of the fds array */
- c->c_fdi = count;
- count++;
- }
+ ct->fd[listnum][count].fd = c->c_prfd;
+ ct->fd[listnum][count].in_flags = SLAPD_POLL_FLAGS;
+ /* slot i of the connection table is mapped to slot
+ * count of the fds array */
+ c->c_fdi = count;
+ count++;
} else {
if (c->c_threadnumber >= c->c_max_threads_per_conn) {
c->c_maxthreadsblocked++;
@@ -1675,7 +1686,7 @@ handle_pr_read_ready(Connection_Table *ct, PRIntn num_poll __attribute__((unused
continue;
}
- /* Try to get connection mutex, if not available just skip the connection and
+ /* Try to get connection mutex, if not available just skip the connection and
* process other connections events. May generates cpu load for listening thread
* if connection mutex is held for a long time
*/
diff --git a/ldap/servers/slapd/opshared.c b/ldap/servers/slapd/opshared.c
index 7ab4117cd..a29eed052 100644
--- a/ldap/servers/slapd/opshared.c
+++ b/ldap/servers/slapd/opshared.c
@@ -250,7 +250,7 @@ op_shared_search(Slapi_PBlock *pb, int send_result)
char *errtext = NULL;
int nentries, pnentries;
int flag_search_base_found = 0;
- int flag_no_such_object = 0;
+ bool flag_no_such_object = false;
int flag_referral = 0;
int flag_psearch = 0;
int err_code = LDAP_SUCCESS;
@@ -315,7 +315,7 @@ op_shared_search(Slapi_PBlock *pb, int send_result)
rc = -1;
goto free_and_return_nolock;
}
-
+
/* Set the time we actually started the operation */
slapi_operation_set_time_started(operation);
@@ -798,11 +798,11 @@ op_shared_search(Slapi_PBlock *pb, int send_result)
}
/* subtree searches :
- * if the search was started above the backend suffix
- * - temporarily set the SLAPI_SEARCH_TARGET_SDN to the
- * base of the node so that we don't get a NO SUCH OBJECT error
- * - do not change the scope
- */
+ * if the search was started above the backend suffix
+ * - temporarily set the SLAPI_SEARCH_TARGET_SDN to the
+ * base of the node so that we don't get a NO SUCH OBJECT error
+ * - do not change the scope
+ */
if (scope == LDAP_SCOPE_SUBTREE) {
if (slapi_sdn_issuffix(be_suffix, basesdn)) {
if (free_sdn) {
@@ -825,53 +825,53 @@ op_shared_search(Slapi_PBlock *pb, int send_result)
switch (rc) {
case 1:
/* if the backend returned LDAP_NO_SUCH_OBJECT for a SEARCH request,
- * it will not have sent back a result - otherwise, it will have
- * sent a result */
+ * it will not have sent back a result - otherwise, it will have
+ * sent a result */
rc = SLAPI_FAIL_GENERAL;
slapi_pblock_get(pb, SLAPI_RESULT_CODE, &err);
if (err == LDAP_NO_SUCH_OBJECT) {
/* may be the object exist somewhere else
- * wait the end of the loop to send back this error
- */
- flag_no_such_object = 1;
+ * wait the end of the loop to send back this error
+ */
+ flag_no_such_object = true;
} else {
/* err something other than LDAP_NO_SUCH_OBJECT, so the backend will
- * have sent the result -
- * Set a flag here so we don't return another result. */
+ * have sent the result -
+ * Set a flag here so we don't return another result. */
sent_result = 1;
}
- /* fall through */
+ /* fall through */
case -1: /* an error occurred */
+ slapi_pblock_get(pb, SLAPI_RESULT_CODE, &err);
/* PAGED RESULTS */
if (op_is_pagedresults(operation)) {
/* cleanup the slot */
pthread_mutex_lock(pagedresults_mutex);
+ if (err != LDAP_NO_SUCH_OBJECT && !flag_no_such_object) {
+ /* Free the results if not "no_such_object" */
+ void *sr = NULL;
+ slapi_pblock_get(pb, SLAPI_SEARCH_RESULT_SET, &sr);
+ be->be_search_results_release(&sr);
+ }
pagedresults_set_search_result(pb_conn, operation, NULL, 1, pr_idx);
rc = pagedresults_set_current_be(pb_conn, NULL, pr_idx, 1);
pthread_mutex_unlock(pagedresults_mutex);
}
- if (1 == flag_no_such_object) {
- break;
- }
- slapi_pblock_get(pb, SLAPI_RESULT_CODE, &err);
- if (err == LDAP_NO_SUCH_OBJECT) {
- /* may be the object exist somewhere else
- * wait the end of the loop to send back this error
- */
- flag_no_such_object = 1;
+
+ if (err == LDAP_NO_SUCH_OBJECT || flag_no_such_object) {
+ /* Maybe the object exists somewhere else, wait to the end
+ * of the loop to send back this error */
+ flag_no_such_object = true;
break;
} else {
- /* for error other than LDAP_NO_SUCH_OBJECT
- * the error has already been sent
- * stop the search here
- */
+ /* For error other than LDAP_NO_SUCH_OBJECT the error has
+ * already been sent stop the search here */
cache_return_target_entry(pb, be, operation);
goto free_and_return;
}
/* when rc == SLAPI_FAIL_DISKFULL this case is executed */
-
case SLAPI_FAIL_DISKFULL:
operation_out_of_disk_space();
cache_return_target_entry(pb, be, operation);
diff --git a/ldap/servers/slapd/pagedresults.c b/ldap/servers/slapd/pagedresults.c
index db87e486e..4aa1fa3e5 100644
--- a/ldap/servers/slapd/pagedresults.c
+++ b/ldap/servers/slapd/pagedresults.c
@@ -121,12 +121,15 @@ pagedresults_parse_control_value(Slapi_PBlock *pb,
if (ber_scanf(ber, "{io}", pagesize, &cookie) == LBER_ERROR) {
slapi_log_err(SLAPI_LOG_ERR, "pagedresults_parse_control_value",
"<= corrupted control value\n");
+ ber_free(ber, 1);
return LDAP_PROTOCOL_ERROR;
}
if (!maxreqs) {
slapi_log_err(SLAPI_LOG_ERR, "pagedresults_parse_control_value",
"Simple paged results requests per conn exceeded the limit: %d\n",
maxreqs);
+ ber_free(ber, 1);
+ slapi_ch_free_string(&cookie.bv_val);
return LDAP_UNWILLING_TO_PERFORM;
}
@@ -376,6 +379,10 @@ pagedresults_free_one_msgid(Connection *conn, ber_int_t msgid, pthread_mutex_t *
}
prp->pr_flags |= CONN_FLAG_PAGEDRESULTS_ABANDONED;
prp->pr_flags &= ~CONN_FLAG_PAGEDRESULTS_PROCESSING;
+ if (conn->c_pagedresults.prl_count > 0) {
+ _pr_cleanup_one_slot(prp);
+ conn->c_pagedresults.prl_count--;
+ }
rc = 0;
break;
}
@@ -940,7 +947,9 @@ pagedresults_is_timedout_nolock(Connection *conn)
return 1;
}
}
+
slapi_log_err(SLAPI_LOG_TRACE, "<-- pagedresults_is_timedout", "<= false 2\n");
+
return 0;
}
diff --git a/ldap/servers/slapd/slap.h b/ldap/servers/slapd/slap.h
index 072f6f962..469874fd1 100644
--- a/ldap/servers/slapd/slap.h
+++ b/ldap/servers/slapd/slap.h
@@ -74,7 +74,7 @@ static char ptokPBE[34] = "Internal (Software) Token ";
#include <sys/stat.h>
#include <sys/socket.h>
#include <netinet/in.h>
-
+#include <stdbool.h>
#include <time.h> /* For timespec definitions */
/* Provides our int types and platform specific requirements. */
--
2.48.0

29
SOURCES/0022-Issue-6509-Fix-cherry-pick-issue-race-condition-in-P.patch Normal file
View File

@ -0,0 +1,29 @@
From 27cd055197bc3cae458a1f86621aa5410c66dd2c Mon Sep 17 00:00:00 2001
From: Mark Reynolds <mreynolds@redhat.com>
Date: Mon, 20 Jan 2025 15:51:24 -0500
Subject: [PATCH] Issue 6509 - Fix cherry pick issue (race condition in Paged
results)
Relates: https://github.com/389ds/389-ds-base/issues/6509
---
ldap/servers/slapd/daemon.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/ldap/servers/slapd/daemon.c b/ldap/servers/slapd/daemon.c
index 13dfe250d..57e07e5f5 100644
--- a/ldap/servers/slapd/daemon.c
+++ b/ldap/servers/slapd/daemon.c
@@ -1620,8 +1620,8 @@ setup_pr_read_pds(Connection_Table *ct)
connection_table_move_connection_out_of_active_list(ct, c);
} else if (c->c_prfd != NULL) {
if ((!c->c_gettingber) && (c->c_threadnumber < c->c_max_threads_per_conn)) {
- ct->fd[listnum][count].fd = c->c_prfd;
- ct->fd[listnum][count].in_flags = SLAPD_POLL_FLAGS;
+ ct->fd[count].fd = c->c_prfd;
+ ct->fd[count].in_flags = SLAPD_POLL_FLAGS;
/* slot i of the connection table is mapped to slot
* count of the fds array */
c->c_fdi = count;
--
2.48.0

8
SPECS/389-ds-base.spec
View File

@ -48,7 +48,7 @@ ExcludeArch: i686
Summary: 389 Directory Server (base) Summary: 389 Directory Server (base)
Name: 389-ds-base Name: 389-ds-base
Version: 1.4.3.39 Version: 1.4.3.39
Release: %{?relprefix}10%{?prerel}%{?dist} Release: %{?relprefix}11%{?prerel}%{?dist}
License: GPLv3+ and (ASL 2.0 or MIT) License: GPLv3+ and (ASL 2.0 or MIT)
URL: https://www.port389.org URL: https://www.port389.org
Group: System Environment/Daemons Group: System Environment/Daemons
@ -313,6 +313,9 @@ Patch17: 0017-Issue-6224-Remove-test_referral_subsuffix-from-ds_lo.patc
Patch18: 0018-Issue-6417-2nd-If-an-entry-RDN-is-identical-to-the-s.patch Patch18: 0018-Issue-6417-2nd-If-an-entry-RDN-is-identical-to-the-s.patch
Patch19: 0019-Issue-6417-2nd-fix-typo.patch Patch19: 0019-Issue-6417-2nd-fix-typo.patch
Patch20: 0020-Issue-6417-3rd-If-an-entry-RDN-is-identical-to-the-s.patch Patch20: 0020-Issue-6417-3rd-If-an-entry-RDN-is-identical-to-the-s.patch
Patch21: 0021-Issue-6509-Race-condition-with-Paged-Result-searches.patch
Patch22: 0022-Issue-6509-Fix-cherry-pick-issue-race-condition-in-P.patch
%description %description
389 Directory Server is an LDAPv3 compliant server. The base package includes 389 Directory Server is an LDAPv3 compliant server. The base package includes
the LDAP server and command line utilities for server administration. the LDAP server and command line utilities for server administration.
@ -933,6 +936,9 @@ exit 0
%doc README.md %doc README.md
%changelog %changelog
* Thu Jan 23 2025 Viktor Ashirov <vashirov@redhat.com> - 1.4.3.39-11
- Resolves: RHEL-72487 - IPA LDAP error code T3 when no exceeded time limit from a paged search result [rhel-8.10.z]
* Fri Jan 17 2025 Viktor Ashirov <vashirov@redhat.com> - 1.4.3.39-10 * Fri Jan 17 2025 Viktor Ashirov <vashirov@redhat.com> - 1.4.3.39-10
- Resolves: RHEL-69822 - "Duplicated DN detected" errors when creating indexes or importing entries. [rhel-8.10.z] - Resolves: RHEL-69822 - "Duplicated DN detected" errors when creating indexes or importing entries. [rhel-8.10.z]
- Resolves: RHEL-71215 - Sub suffix causes "id2entry - Could not open id2entry err 0" error when the Directory Server starts [rhel-8.10.z] - Resolves: RHEL-71215 - Sub suffix causes "id2entry - Could not open id2entry err 0" error when the Directory Server starts [rhel-8.10.z]