In ssl3_write_bytes, some checks related to hanlding write failure are missing

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

In ssl3_write_bytes, some checks related to hanlding write failure are missing

Raja ashok

Hi,

 

I feel there is a check missing in ssl3_write_bytes, in case of handling write failure.

 

Consider SSL_write is called with 20000 bytes buffer, then internally in ssl3_write_bytes we try to send it as two record (16384 and 3616). If TCP send failed for the second record then we store the states internally (wnum, wpend_tot and wpend_buf) and return back the result.

 

Later application has to call SSL_write with same buffer, if it calls with different buffer of length 100 byte then we fail that in ssl3_write_bytes using the check (len < tot).

 

But consider application calls with buffer of size 18000 bytes and SSL_MODE_ACCEPT_MOVING_WRITE_BUFFER is enabled. Then (len < tot) will not succeed as tot is 16384. Then it will call ssl3_write_pending to send the remaining 3616 record. If it succeeds we are incrementing tot, (tot += i). Now tot will have 20000.

 

Later there is a check (tot == len), this will not succeed. Then directly we are doing n = (len - tot), this will overflow and store a value close to 2^32 in n. Then it will cause out of bound access to the application buffer "buf".

 

I hope we should have one more check (len < (tot + s->s3->wpend_tot)) before calling ssl3_write_pending.

 

    if ((len < tot) || (len < (tot + s->s3->wpend_tot))){

        SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);

        return (-1);

}

 

Note : I am referring 1.0.2k version of OpenSSL.

 

Regards,

Ashok

 


Company_logo

Raja Ashok V K
Huawei Technologies
Bangalore, India
http://www.huawei.com


本邮件及其附件含有华为公司的保密信息,仅限于发送给上面地址中列出的个人或群组。禁
止任何其他人以任何形式使用(包括但不限于全部或部分地泄露、复制、或散发)本邮件中
的信息。如果您错收了本邮件,请您立即电话或邮件通知发件人并删除本邮件!
This e-mail and its attachments contain confidential information from HUAWEI, which
is intended only for the person or entity whose address is listed above. Any use of the
information contained herein in any way (including, but not limited to, total or partial
disclosure, reproduction, or dissemination) by persons other than the intended
recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by
phone or email immediately and delete it!

 


--
openssl-dev mailing list
To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: In ssl3_write_bytes, some checks related to hanlding write failure are missing

Raja ashok

Hi All,

 

In ssl3_write_bytes, if (len < tot) we are returning failure with SSL_R_BAD_LENGTH error. In this place I hope we should set “tot” back to “s->s3->wnum”. Otherwise when application calls back SSL_write with correct buffer, it causes serious problem (“tot” is 0 and iLeft is not NULL). I hope we should do like below.

 

    if (len < tot) {

        s->s3->wnum = tot;

        SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);

       return (-1);

    }

 

And also we should do one additional check for “len” as mentioned in my previous mail.

 

    if ((len < tot) || ((tot != 0) && (len < (tot + s->s3->wpend_tot)))){

        s->s3->wnum = tot;

        SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);

       return (-1);

    }

 

Regards,

Ashok

 


Company_logo

Raja Ashok V K
Huawei Technologies
Bangalore, India
http://www.huawei.com


本邮件及其附件含有华为公司的保密信息,仅限于发送给上面地址中列出的个人或群组。禁
止任何其他人以任何形式使用(包括但不限于全部或部分地泄露、复制、或散发)本邮件中
的信息。如果您错收了本邮件,请您立即电话或邮件通知发件人并删除本邮件!
This e-mail and its attachments contain confidential information from HUAWEI, which
is intended only for the person or entity whose address is listed above. Any use of the
information contained herein in any way (including, but not limited to, total or partial
disclosure, reproduction, or dissemination) by persons other than the intended
recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by
phone or email immediately and delete it!

 

From: Raja ashok
Sent: 27 March 2017 13:55
To: '[hidden email]'; '[hidden email]'
Subject: In ssl3_write_bytes, some checks related to hanlding write failure are missing

 

Hi,

 

I feel there is a check missing in ssl3_write_bytes, in case of handling write failure.

 

Consider SSL_write is called with 20000 bytes buffer, then internally in ssl3_write_bytes we try to send it as two record (16384 and 3616). If TCP send failed for the second record then we store the states internally (wnum, wpend_tot and wpend_buf) and return back the result.

 

Later application has to call SSL_write with same buffer, if it calls with different buffer of length 100 byte then we fail that in ssl3_write_bytes using the check (len < tot).

 

But consider application calls with buffer of size 18000 bytes and SSL_MODE_ACCEPT_MOVING_WRITE_BUFFER is enabled. Then (len < tot) will not succeed as tot is 16384. Then it will call ssl3_write_pending to send the remaining 3616 record. If it succeeds we are incrementing tot, (tot += i). Now tot will have 20000.

 

Later there is a check (tot == len), this will not succeed. Then directly we are doing n = (len - tot), this will overflow and store a value close to 2^32 in n. Then it will cause out of bound access to the application buffer "buf".

 

I hope we should have one more check (len < (tot + s->s3->wpend_tot)) before calling ssl3_write_pending.

 

    if ((len < tot) || (len < (tot + s->s3->wpend_tot))){

        SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);

        return (-1);

}

 

Note : I am referring 1.0.2k version of OpenSSL.

 

Regards,

Ashok

 


Company_logo

Raja Ashok V K
Huawei Technologies
Bangalore, India
http://www.huawei.com


本邮件及其附件含有华为公司的保密信息,仅限于发送给上面地址中列出的个人或群组。禁
止任何其他人以任何形式使用(包括但不限于全部或部分地泄露、复制、或散发)本邮件中
的信息。如果您错收了本邮件,请您立即电话或邮件通知发件人并删除本邮件!
This e-mail and its attachments contain confidential information from HUAWEI, which
is intended only for the person or entity whose address is listed above. Any use of the
information contained herein in any way (including, but not limited to, total or partial
disclosure, reproduction, or dissemination) by persons other than the intended
recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by
phone or email immediately and delete it!

 


--
openssl-dev mailing list
To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: In ssl3_write_bytes, some checks related to hanlding write failure are missing

Matt Caswell-2
On 31/03/17 18:54, Raja ashok wrote:

> Hi All,
>
>
>
> In ssl3_write_bytes, if (len < tot) we are returning failure with
> SSL_R_BAD_LENGTH error. In this place I hope we should set “tot” back to
> “s->s3->wnum”. Otherwise when application calls back SSL_write with
> correct buffer, it causes serious problem (“tot” is 0 and iLeft is not
> NULL). I hope we should do like below.
>
>
>
>     if (len < tot) {
>
>         s->s3->wnum = tot;
>
>         SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);
>
>        return (-1);
>
>     }

This is 1.0.2 code. The check appears to be earlier in master/1.1.0
(before wnum is reset) and so this isn't an issue there. Really, if an
application passes a bad len value, then this is an application bug and
shouldn't ever happen in a well-behaved application. I'm not sure you
could really describe this as an OpenSSL bug (its a bit border line) so
I'm not sure it justifies a patch to 1.0.2 (which only takes bug fixes).

>
> And also we should do one additional check for “len” as mentioned in my
> previous mail.
>
>
>
>     if ((len < tot) || ((tot != 0) && (len < (tot + s->s3->wpend_tot)))){
>
>         s->s3->wnum = tot;
>
>         SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);
>
>        return (-1);
>
>     }

Please could you raise a github pull request for this suggestion? You
will probably need two versions: one targeting master and one targeting
1.0.2 as the the code looks a little different in this area.

Matt
--
openssl-dev mailing list
To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: In ssl3_write_bytes, some checks related to hanlding write failure are missing

Raja ashok
Hi Matt,

I created a Pull request for the second change on master, 1.0.2 and 1.1.0. I am creating PR in github for the first time, so if anything else I missed please update me.

https://github.com/openssl/openssl/pull/3124
https://github.com/openssl/openssl/pull/3123
https://github.com/openssl/openssl/pull/3122

Regards,
Ashok



Raja Ashok V K
Huawei Technologies
Bangalore, India
http://www.huawei.com 

本邮件及其附件含有华为公司的保密信息,仅限于发送给上面地址中列出的个人或群组。禁
止任何其他人以任何形式使用(包括但不限于全部或部分地泄露、复制、或散发)本邮件中
的信息。如果您错收了本邮件,请您立即电话或邮件通知发件人并删除本邮件!
This e-mail and its attachments contain confidential information from HUAWEI, which
is intended only for the person or entity whose address is listed above. Any use of the
information contained herein in any way (including, but not limited to, total or partial
disclosure, reproduction, or dissemination) by persons other than the intended
recipient(s) is prohibited. If you receive this e-mail in error, please notify the sender by
phone or email immediately and delete it!


-----Original Message-----
From: openssl-dev [mailto:[hidden email]] On Behalf Of Matt Caswell
Sent: 03 April 2017 14:40
To: [hidden email]
Subject: Re: [openssl-dev] In ssl3_write_bytes, some checks related to hanlding write failure are missing

On 31/03/17 18:54, Raja ashok wrote:

> Hi All,
>
>
>
> In ssl3_write_bytes, if (len < tot) we are returning failure with
> SSL_R_BAD_LENGTH error. In this place I hope we should set “tot” back
> to “s->s3->wnum”. Otherwise when application calls back SSL_write with
> correct buffer, it causes serious problem (“tot” is 0 and iLeft is not
> NULL). I hope we should do like below.
>
>
>
>     if (len < tot) {
>
>         s->s3->wnum = tot;
>
>         SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);
>
>        return (-1);
>
>     }

This is 1.0.2 code. The check appears to be earlier in master/1.1.0 (before wnum is reset) and so this isn't an issue there. Really, if an application passes a bad len value, then this is an application bug and shouldn't ever happen in a well-behaved application. I'm not sure you could really describe this as an OpenSSL bug (its a bit border line) so I'm not sure it justifies a patch to 1.0.2 (which only takes bug fixes).

>
> And also we should do one additional check for “len” as mentioned in
> my previous mail.
>
>
>
>     if ((len < tot) || ((tot != 0) && (len < (tot +
> s->s3->wpend_tot)))){
>
>         s->s3->wnum = tot;
>
>         SSLerr(SSL_F_SSL3_WRITE_BYTES, SSL_R_BAD_LENGTH);
>
>        return (-1);
>
>     }

Please could you raise a github pull request for this suggestion? You will probably need two versions: one targeting master and one targeting
1.0.2 as the the code looks a little different in this area.

Matt
--
openssl-dev mailing list
To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
--
openssl-dev mailing list
To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Loading...