freebsd security - Jan 2018 - Need FreeBSD-SA-00:52(TCP uses weak initial sequence numbers) latest patch

On 2018-Jan-12 12:33:21 +0530, Brahmanand Reddy <brahma.gdb at gmail.com>
wrote:
>TCP uses weak initial sequence numbers
>https://www.freebsd.org/security/advisories/FreeBSD-SA-00%3A52.tcp-iss.asc
As has been pointed out to you several times in this thread, that SA is
nearly 20 years old and there is no evidence that TCP on any recent FreeBSD
uses weak ISNs.
>actually "arc4random()"  will take care on 
https://github.com/freebsd/
>freebsd/blob/master/sys/netinet/tcp_subr.c#L2374
Without studying the code in detail, that code appears to correctly use
arc4random() to initialise the ISN - which is as expected.
> I suspecting 10.4 already having fix... but i didn't found on exactly
>which this problem from  https://www.freebsd.org/security/patches/
Well, the original patch is
https://www.freebsd.org/security/patches/SA-00%3A52/ and was committed
as what is now https://svnweb.freebsd.org/base?view=revision&revision=66433
Since that patch is integrated into the FreeBSD codebase, there's no need
to update the contents of https://www.freebsd.org/security/patches/SA-00%3A52/
and it is not relevant to the current codebase.
>  i would like expecting where is the fix in 10,4 kernel.
That code was re-written in r82122, retaining the use of arc4random() for
ISN initialisation.  As a result, it's no longer possible to point at
specific code and say "that code fixes weak TCP ISNs".

-- 
Peter Jeremy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 963 bytes
Desc: not available
URL:
<http://lists.freebsd.org/pipermail/freebsd-security/attachments/20180112/0eb203de/attachment.sig>

Hi Peter,

My last question on this ,  recently  "Replaced the kernel RC4(arc4random)
with Chacha20" on 11.0 kernel should we apply on 10.4 kernel ??
 please find the corresponding review and fix
https://reviews.freebsd.org/D10048 and  https://reviews.freebsd.org/rS317015


Thanks in advance,
Brahma

On Fri, Jan 12, 2018 at 1:11 PM, Peter Jeremy <peter at rulingia.com>
wrote:
> On 2018-Jan-12 12:33:21 +0530, Brahmanand Reddy <brahma.gdb at
gmail.com>
> wrote:
> >TCP uses weak initial sequence numbers
> >https://www.freebsd.org/security/advisories/FreeBSD-
> SA-00%3A52.tcp-iss.asc
>
> As has been pointed out to you several times in this thread, that SA is
> nearly 20 years old and there is no evidence that TCP on any recent FreeBSD
> uses weak ISNs.
>
> >actually "arc4random()"  will take care on 
https://github.com/freebsd/
> >freebsd/blob/master/sys/netinet/tcp_subr.c#L2374
>
> Without studying the code in detail, that code appears to correctly use
> arc4random() to initialise the ISN - which is as expected.
>
> > I suspecting 10.4 already having fix... but i didn't found on
exactly
> >which this problem from  https://www.freebsd.org/security/patches/
>
> Well, the original patch is
> https://www.freebsd.org/security/patches/SA-00%3A52/ and was committed
> as what is now
https://svnweb.freebsd.org/base?view=revision&revision> 66433
> Since that patch is integrated into the FreeBSD codebase, there's no
need
> to update the contents of https://www.freebsd.org/
> security/patches/SA-00%3A52/
> and it is not relevant to the current codebase.
>
> >  i would like expecting where is the fix in 10,4 kernel.
>
> That code was re-written in r82122, retaining the use of arc4random() for
> ISN initialisation.  As a result, it's no longer possible to point at
> specific code and say "that code fixes weak TCP ISNs".
>
> --
> Peter Jeremy
>

Dear Peter/ Team,

My final call on this thread, "RST not happens as quickly in 10.4 and 11
FreeBSD,  like 9.2".  it takes 10 to 15 seconds delay to reset and generate
new ISN number on 80/443/ports.

  Example-
  # RST
RST=TCP(sport=sport, dport=dport, flags='R', seq=SYNACK.ack, ack=0)
send(ip/RST)
* time.sleep(15*)
SYN2=TCP(sport=sport,dport=dport,flags='S',seq=random.randint(1024,18576),
ack=0)

Kindly clarify and conclude, 1) it could be the expected behavior or any
new enhancement causes this delay what are those.?
                                         2) we have to be configure any
thing like in 'sysctl.conf 'or corresponding fix is available ?

 Note: In 9.2 with out delay . i dint observed any issue. the standards say
it must respond quickly with a RST.

Sincerely,
Brahma



On Fri, Jan 12, 2018 at 1:11 PM, Peter Jeremy <peter at rulingia.com>
wrote:
> On 2018-Jan-12 12:33:21 +0530, Brahmanand Reddy <brahma.gdb at
gmail.com>
> wrote:
> >TCP uses weak initial sequence numbers
> >https://www.freebsd.org/security/advisories/FreeBSD-
> SA-00%3A52.tcp-iss.asc
>
> As has been pointed out to you several times in this thread, that SA is
> nearly 20 years old and there is no evidence that TCP on any recent FreeBSD
> uses weak ISNs.
>
> >actually "arc4random()"  will take care on 
https://github.com/freebsd/
> >freebsd/blob/master/sys/netinet/tcp_subr.c#L2374
>
> Without studying the code in detail, that code appears to correctly use
> arc4random() to initialise the ISN - which is as expected.
>
> > I suspecting 10.4 already having fix... but i didn't found on
exactly
> >which this problem from  https://www.freebsd.org/security/patches/
>
> Well, the original patch is
> https://www.freebsd.org/security/patches/SA-00%3A52/ and was committed
> as what is now
https://svnweb.freebsd.org/base?view=revision&revision> 66433
> Since that patch is integrated into the FreeBSD codebase, there's no
need
> to update the contents of https://www.freebsd.org/
> security/patches/SA-00%3A52/
> and it is not relevant to the current codebase.
>
> >  i would like expecting where is the fix in 10,4 kernel.
>
> That code was re-written in r82122, retaining the use of arc4random() for
> ISN initialisation.  As a result, it's no longer possible to point at
> specific code and say "that code fixes weak TCP ISNs".
>
> --
> Peter Jeremy
>