rpms/krb5
7
0
Fork 0
Code Issues Pull Requests Releases Activity

Add proposed fix for handling AS client clock skew

Browse Source 
In addition to basing the contents of an encrypted-timestamp preauth
data item on the server's idea of the current time, go ahead and do the
same for the times in the request.
This commit is contained in:
Nalin Dahyabhai 2013-05-28 17:47:00 -04:00
parent 9b1daaef43
commit e98d94d2bc
3 changed files with 267 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

115
krb5-1.11.2-skew1.patch Normal file
View File

@ -0,0 +1,115 @@
>From 8f6d12bae1a0f1d274593c4a06dfa5948aa61418 Mon Sep 17 00:00:00 2001
From: Stef Walter <stefw@redhat.com>
Date: Thu, 23 May 2013 08:38:20 +0200
Subject: [PATCH 1/2] krb5: Refator duplicate code for setting the AS REQ nonce
---
src/lib/krb5/krb/get_in_tkt.c | 64 +++++++++++++++++++++++--------------------
1 file changed, 35 insertions(+), 29 deletions(-)
diff --git a/src/lib/krb5/krb/get_in_tkt.c b/src/lib/krb5/krb/get_in_tkt.c
index 828b0fb..1058112 100644
--- a/src/lib/krb5/krb/get_in_tkt.c
+++ b/src/lib/krb5/krb/get_in_tkt.c
@@ -650,6 +650,34 @@ cleanup:
return code;
}
+static krb5_error_code
+update_req_before_encoding(krb5_context context, krb5_init_creds_context ctx)
+{
+ krb5_error_code code = 0;
+ unsigned char random_buf[4];
+ krb5_data random_data;
+
+ /*
+ * RFC 6113 requires a new nonce for the inner request on each try. It's
+ * permitted to change the nonce even for non-FAST as well.
+ */
+ random_data.length = 4;
+ random_data.data = (char *)random_buf;
+ code = krb5_c_random_make_octets(context, &random_data);
+ if (code != 0)
+ goto cleanup;
+
+ /*
+ * See RT ticket 3196 at MIT. If we set the high bit, we may have
+ * compatibility problems with Heimdal, because we (incorrectly) encode
+ * this value as signed.
+ */
+ ctx->request->nonce = 0x7fffffff & load_32_n(random_buf);
+
+cleanup:
+ return code;
+}
+
/**
* Throw away any state related to specific realm either at the beginning of a
* request, or when a realm changes, or when we start to use FAST after
@@ -664,8 +692,6 @@ restart_init_creds_loop(krb5_context context, krb5_init_creds_context ctx,
krb5_pa_data **padata)
{
krb5_error_code code = 0;
- unsigned char random_buf[4];
- krb5_data random_data;
krb5_timestamp from;
if (ctx->preauth_to_use) {
@@ -693,18 +719,10 @@ restart_init_creds_loop(krb5_context context, krb5_init_creds_context ctx,
goto cleanup;
}
- /* Set the request nonce. */
- random_data.length = 4;
- random_data.data = (char *)random_buf;
- code = krb5_c_random_make_octets(context, &random_data);
- if (code !=0)
+ code = update_req_before_encoding(context, ctx);
+ if (code != 0)
goto cleanup;
- /*
- * See RT ticket 3196 at MIT. If we set the high bit, we may have
- * compatibility problems with Heimdal, because we (incorrectly) encode
- * this value as signed.
- */
- ctx->request->nonce = 0x7fffffff & load_32_n(random_buf);
+
krb5_free_principal(context, ctx->request->server);
ctx->request->server = NULL;
@@ -1188,28 +1206,16 @@ init_creds_step_request(krb5_context context,
{
krb5_error_code code;
krb5_boolean got_real;
- char random_buf[4];
- krb5_data random_data;
if (ctx->loopcount >= MAX_IN_TKT_LOOPS) {
code = KRB5_GET_IN_TKT_LOOP;
goto cleanup;
}
- /*
- * RFC 6113 requires a new nonce for the inner request on each try. It's
- * permitted to change the nonce even for non-FAST so we do here.
- */
- random_data.length = 4;
- random_data.data = (char *)random_buf;
- code = krb5_c_random_make_octets(context, &random_data);
- if (code !=0)
+
+ code = update_req_before_encoding(context, ctx);
+ if (code != 0)
goto cleanup;
- /*
- * See RT ticket 3196 at MIT. If we set the high bit, we may have
- * compatibility problems with Heimdal, because we (incorrectly) encode
- * this value as signed.
- */
- ctx->request->nonce = 0x7fffffff & load_32_n(random_buf);
+
krb5_free_data(context, ctx->inner_request_body);
ctx->inner_request_body = NULL;
code = encode_krb5_kdc_req_body(ctx->request, &ctx->inner_request_body);
--
1.8.1.4

144
krb5-1.11.2-skew2.patch Normal file
View File

@ -0,0 +1,144 @@
>From 51ab359d7cc6643cfd4fac28def2e1c756553201 Mon Sep 17 00:00:00 2001
From: Stef Walter <stefw@redhat.com>
Date: Thu, 23 May 2013 08:44:43 +0200
Subject: [PATCH 2/2] krb5: Fix ticket start and end time to respect skew
Since the kerberos protocol uses timestamp rather than duration deltas
for its starttime, endtime, and renewtime KDC AS REQ fields, we have
to calculate these with respect to the offsets we know about received
from the server.
Leverage the unauthenticated server time we received during preauth when
calculating these these timestamps from the duration deltas we use
in our krb5 api and tools.
In order to do this we have to update certain fields of the AS REQ
each time we encode it for sending to the KDC.
---
src/lib/krb5/krb/get_in_tkt.c | 44 +++++++++++++++++++++++--------------------
src/lib/krb5/krb/int-proto.h | 5 +++++
src/lib/krb5/krb/preauth2.c | 8 ++++++++
3 files changed, 37 insertions(+), 20 deletions(-)
diff --git a/src/lib/krb5/krb/get_in_tkt.c b/src/lib/krb5/krb/get_in_tkt.c
index 1058112..694c9b0b 100644
--- a/src/lib/krb5/krb/get_in_tkt.c
+++ b/src/lib/krb5/krb/get_in_tkt.c
@@ -656,6 +656,8 @@ update_req_before_encoding(krb5_context context, krb5_init_creds_context ctx)
krb5_error_code code = 0;
unsigned char random_buf[4];
krb5_data random_data;
+ krb5_timestamp from;
+ krb5_int32 unused;
/*
* RFC 6113 requires a new nonce for the inner request on each try. It's
@@ -674,6 +676,28 @@ update_req_before_encoding(krb5_context context, krb5_init_creds_context ctx)
*/
ctx->request->nonce = 0x7fffffff & load_32_n(random_buf);
+ code = k5_preauth_get_time(context, &ctx->preauth_rock, TRUE, &ctx->request_time, &unused);
+ if (code != 0)
+ goto cleanup;
+
+ /* Omit request start time in the common case. MIT and Heimdal KDCs will
+ * ignore it for non-postdated tickets anyway. */
+ from = krb5int_addint32(ctx->request_time, ctx->start_time);
+ if (ctx->start_time != 0)
+ ctx->request->from = from;
+ ctx->request->till = krb5int_addint32(from, ctx->tkt_life);
+
+ if (ctx->renew_life > 0) {
+ ctx->request->rtime =
+ krb5int_addint32(from, ctx->renew_life);
+ if (ctx->request->rtime < ctx->request->till) {
+ /* don't ask for a smaller renewable time than the lifetime */
+ ctx->request->rtime = ctx->request->till;
+ }
+ ctx->request->kdc_options &= ~(KDC_OPT_RENEWABLE_OK);
+ } else
+ ctx->request->rtime = 0;
+
cleanup:
return code;
}
@@ -692,7 +716,6 @@ restart_init_creds_loop(krb5_context context, krb5_init_creds_context ctx,
krb5_pa_data **padata)
{
krb5_error_code code = 0;
- krb5_timestamp from;
if (ctx->preauth_to_use) {
krb5_free_pa_data(context, ctx->preauth_to_use);
@@ -732,8 +755,6 @@ restart_init_creds_loop(krb5_context context, krb5_init_creds_context ctx,
if (code != 0)
goto cleanup;
- ctx->request_time = time(NULL);
-
code = krb5int_fast_as_armor(context, ctx->fast_state,
ctx->opte, ctx->request);
if (code != 0)
@@ -747,23 +768,6 @@ restart_init_creds_loop(krb5_context context, krb5_init_creds_context ctx,
/* give the preauth plugins a chance to prep the request body */
krb5_preauth_prepare_request(context, ctx->opte, ctx->request);
- /* Omit request start time in the common case. MIT and Heimdal KDCs will
- * ignore it for non-postdated tickets anyway. */
- from = krb5int_addint32(ctx->request_time, ctx->start_time);
- if (ctx->start_time != 0)
- ctx->request->from = from;
- ctx->request->till = krb5int_addint32(from, ctx->tkt_life);
-
- if (ctx->renew_life > 0) {
- ctx->request->rtime =
- krb5int_addint32(from, ctx->renew_life);
- if (ctx->request->rtime < ctx->request->till) {
- /* don't ask for a smaller renewable time than the lifetime */
- ctx->request->rtime = ctx->request->till;
- }
- ctx->request->kdc_options &= ~(KDC_OPT_RENEWABLE_OK);
- } else
- ctx->request->rtime = 0;
code = krb5int_fast_prep_req_body(context, ctx->fast_state,
ctx->request,
&ctx->outer_request_body);
diff --git a/src/lib/krb5/krb/int-proto.h b/src/lib/krb5/krb/int-proto.h
index 3326154..83a47c0 100644
--- a/src/lib/krb5/krb/int-proto.h
+++ b/src/lib/krb5/krb/int-proto.h
@@ -142,6 +142,11 @@ krb5_preauth_supply_preauth_data(krb5_context context,
const char *value);
krb5_error_code
+k5_preauth_get_time(krb5_context context, krb5_clpreauth_rock rock,
+ krb5_boolean allow_unauth_time, krb5_timestamp *time_out,
+ krb5_int32 *usec_out);
+
+krb5_error_code
clpreauth_encrypted_challenge_initvt(krb5_context context, int maj_ver,
int min_ver, krb5_plugin_vtable vtable);
diff --git a/src/lib/krb5/krb/preauth2.c b/src/lib/krb5/krb/preauth2.c
index 747611e..167f611 100644
--- a/src/lib/krb5/krb/preauth2.c
+++ b/src/lib/krb5/krb/preauth2.c
@@ -397,6 +397,15 @@ get_preauth_time(krb5_context context, krb5_clpreauth_rock rock,
krb5_boolean allow_unauth_time, krb5_timestamp *time_out,
krb5_int32 *usec_out)
{
+ return k5_preauth_get_time(context, rock, allow_unauth_time,
+ time_out, usec_out);
+}
+
+krb5_error_code
+k5_preauth_get_time(krb5_context context, krb5_clpreauth_rock rock,
+ krb5_boolean allow_unauth_time, krb5_timestamp *time_out,
+ krb5_int32 *usec_out)
+{
if (rock->pa_offset_state != NO_OFFSET &&
(allow_unauth_time || rock->pa_offset_state == AUTH_OFFSET) &&
(context->library_options & KRB5_LIBOPT_SYNC_KDCTIME)) {
--
1.8.1.4

8
krb5.spec
View File

@ -82,6 +82,8 @@ Patch121: krb5-cccol-primary.patch
Patch122: krb5-1.11.2-gss_transited.patch
Patch123: krb5-1.11.2-empty_passwords.patch
Patch124: krb5-1.11.2-arcfour_short.patch
Patch125: krb5-1.11.2-skew1.patch
Patch126: krb5-1.11.2-skew2.patch
# Patches for otp plugin backport
Patch201: krb5-1.11.2-keycheck.patch
@ -306,6 +308,8 @@ ln -s NOTICE LICENSE
%patch122 -p1 -b .gss_transited
%patch123 -p1 -b .empty_passwords
%patch124 -p1 -b .arcfour_short
%patch125 -p1 -b .skew1
%patch126 -p1 -b .skew2
%patch201 -p1 -b .keycheck
%patch202 -p1 -b .otp
@ -836,6 +840,10 @@ exit 0
in GSS acceptors (RT#7639, #959685)
- backport fix for not being able to pass an empty password to the
get-init-creds APIs and have them actually use it (RT#7642, #960001)
- add backported proposed fix to use the unauthenticated server time
as the basis for computing the requested credential expiration times,
rather than the client's idea of the current time, which could be
significantly incorrect (#961221)
* Tue May 21 2013 Nalin Dahyabhai <nalin@redhat.com> 1.11.2-6
- pull in upstream fix to start treating a KRB5CCNAME value that begins