[PATCH] Avoid using the result of i2d_SSL_SESSION when the session is invalid
Bart Warmerdam
bartw at xs4all.nl
Mon Jun 19 14:09:43 UTC 2017
According to the man-page of i2d_SSL_SESSION the result can be NULL or
0, but case the actual result can also be -1 in case of a failed
CRYPTO_malloc. The call trace for this function is:
Call chain:
i2d_SSL_SESSION
i2d_SSL_SESSION_ASN1
ASN1_item_i2d
asn1_item_flags_i2d
The preprocessor output generates the following code:
static int asn1_item_flags_i2d(ASN1_VALUE *val, unsigned char **out,
const ASN1_ITEM *it, int flags)
{
if (out && !*out) {
unsigned char *p, *buf;
int len;
len = ASN1_item_ex_i2d(&val,
# 60 "crypto/asn1/tasn_enc.c" 3 4
((void *)0)
# 60 "crypto/asn1/tasn_enc.c"
, it, -1, flags);
if (len <= 0)
return len;
buf = CRYPTO_malloc(len, "crypto/asn1/tasn_enc.c", 63);
if (buf ==
# 64 "crypto/asn1/tasn_enc.c" 3 4
((void *)0)
# 64 "crypto/asn1/tasn_enc.c"
)
return -1;
p = buf;
ASN1_item_ex_i2d(&val, &p, it, -1, flags);
*out = buf;
return len;
}
return ASN1_item_ex_i2d(&val, out, it, -1, flags);
}
As you can see around line 65 the -1 is returned once the allocation
fails, which is then directly returned in all intermediate calls up to
i2d_SSL_SESSION. So the -1 case is missed once the current "if
statement" is kept without the "len < 1" or-statement.
Regards,
B.
On 2017-06-19 12:10, Ruslan Ermilov wrote:
> On Mon, Jun 19, 2017 at 08:08:38AM +0200, Bart Warmerdam wrote:
>> # HG changeset patch
>> # User Bart Warmerdam <bartw at xs4all.nl>
>> # Date 1497852211 -7200
>> # Mon Jun 19 08:03:31 2017 +0200
>> # Branch i2d_ssl_session_length
>> # Node ID 079afb2cb4be3ef06d07e96d1a54cc359b971631
>> # Parent d1816a2696de8c2faa1cd913a151e5f62a8620f3
>> Make sure to also take into account the 'return 0' response of
>> i2d_SSL_SESSION, which is possible when the session is not valid
>
> A case of invalid session is already caught by checking the return
> value of SSL_get0_session() just prior to calling i2d_SSL_SESSION()
> in ngx_http_upstream_save_round_robin_peer_session() and
> ngx_stream_upstream_save_round_robin_peer_session().
>
> The ngx_ssl_new_session() function is passed as an argument to
> SSL_CTX_sess_set_new_cb() that "sets the callback function, which
> is automatically called whenever a new session was negotiated."
>
> Do you think that it's possible for a session to be invalid in
> either of these cases?
>
>> diff -r d1816a2696de -r 079afb2cb4be src/event/ngx_event_openssl.c
>> --- a/src/event/ngx_event_openssl.c Fri Jun 16 18:15:58 2017 +0300
>> +++ b/src/event/ngx_event_openssl.c Mon Jun 19 08:03:31 2017 +0200
>> @@ -2458,9 +2458,9 @@
>>
>> len = i2d_SSL_SESSION(sess, NULL);
>>
>> - /* do not cache too big session */
>> -
>> - if (len > (int) NGX_SSL_MAX_SESSION_SIZE) {
>> + /* do not cache too big or invalid session */
>> +
>> + if (len > (int) NGX_SSL_MAX_SESSION_SIZE || len < 1) {
>> return 0;
>> }
>>
>> diff -r d1816a2696de -r 079afb2cb4be
>> src/http/ngx_http_upstream_round_robin.c
>> --- a/src/http/ngx_http_upstream_round_robin.c Fri Jun 16 18:15:58
>> 2017
>> +0300
>> +++ b/src/http/ngx_http_upstream_round_robin.c Mon Jun 19 08:03:31
>> 2017
>> +0200
>> @@ -755,9 +755,9 @@
>>
>> len = i2d_SSL_SESSION(ssl_session, NULL);
>>
>> - /* do not cache too big session */
>> + /* do not cache too big or invalid session */
>>
>> - if (len > NGX_SSL_MAX_SESSION_SIZE) {
>> + if (len > NGX_SSL_MAX_SESSION_SIZE || len < 1) {
>> return;
>> }
>>
>> diff -r d1816a2696de -r 079afb2cb4be
>> src/stream/ngx_stream_upstream_round_robin.c
>> --- a/src/stream/ngx_stream_upstream_round_robin.c Fri Jun 16
>> 18:15:58 2017 +0300
>> +++ b/src/stream/ngx_stream_upstream_round_robin.c Mon Jun 19
>> 08:03:31 2017 +0200
>> @@ -787,9 +787,9 @@
>>
>> len = i2d_SSL_SESSION(ssl_session, NULL);
>>
>> - /* do not cache too big session */
>> + /* do not cache too big or invalid session */
>>
>> - if (len > NGX_SSL_MAX_SESSION_SIZE) {
>> + if (len > NGX_SSL_MAX_SESSION_SIZE || len < 1) {
>> return;
>> }
>> _______________________________________________
>> nginx-devel mailing list
>> nginx-devel at nginx.org
>> http://mailman.nginx.org/mailman/listinfo/nginx-devel
>>
More information about the nginx-devel
mailing list