_ ___ ____ ____ ____ _ _ |_|_ _ / _ \/ ___/ ___|| _ \ _ _ _ _(_) __| | _|_||_| | | | \___ \___ \| |_) | | | | | | | | |/ _` | |_||_|_| | |_| |___) |__) | __/ | |_| | |_| | | (_| | |_|_|_| \___/|____/____/|_| \__,_|\__,_|_|\__,_| OSSP uuid - Universally Unique Identifier HISTORY During OSSP uuid we were totally puzzled by a subtle bug in the UUID standards related to the generation of multi-cast MAC addresses. This part of the history shows a very interesting technical bug, the unusual way of having to fix a standard (which was multiple times revised by different standard authorities, including the IETF, the OpenGroup and ISO/IEC) afterwards plus the fixing of six implementations into which the bug was inherited similarly. Below are some snapshot of this part of history: the first implementation fix (for FreeBSD) and the notification of the IETF standards authors. ___________________________________________________________________________ Date: Fri, 13 Feb 2004 16:09:31 +0100 From: "Ralf S. Engelschall" To: paulle@microsoft.com, michael@neonym.net, rsalz@datapower.com Subject: [PATCH] draft-mealling-uuid-urn-02.txt Message-ID: <20040213150931.GA7656@engelschall.com> During implementation of OSSP uuid (a flexible CLI and C API for generation and partial decoding of version 1, 3 and 4 UUIDs, see http://www.ossp.org/pkg/lib/uuid/ for details), I discovered a nasty bug in the generation of random multicast MAC addresses. It is present in all standards and drafts (both expired ones and current ones) and was also inherited (until I fixed it by submitting patches to the authors recently) by all six freely available UUID implementations (Apache APR, FreeBSD uuidgen(2), Java JUG, Linux's libuuid from e2fsutil, Perl's Data::UUID and WINE's UUID generator)). In case no real/physical IEEE 802 address is available, both the expired "draft-leach-uuids-guids-01" (section "4. Node IDs when no IEEE 802 network card is available"), RFC 2518 (section "6.4.1 Node Field Generation Without the IEEE 802 Address") and now even your current "draft-mealling-uuid-urn-02.txt" (section "4.5 Node IDs that do not identify the host") recommend: "A better solution is to obtain a 47-bit cryptographic quality random number, and use it as the low 47 bits of the node ID, with the _most_ significant bit of the first octet of the node ID set to one. This bit is the unicast/multicast bit, which will never be set in IEEE 802 addresses obtained from network cards; hence, there can never be a conflict between UUIDs generated by machines with and without network cards." Unfortunately, this incorrectly explains how to implement this and even the example implementation (draft-mealling-uuid-urn-02.txt, "Appendix A. Appendix A - Sample Implementation") inherited this. Correct is "the _least_ significant bit of the first octet of the node ID" as the multicast bit in a memory and hexadecimal string representation of a 48-bit IEEE 802 MAC address. This standards bug arised from a false interpretation, as the multicast bit is actually the _most_ significant bit in IEEE 802.3 (Ethernet) _transmission order_ of an IEEE 802 MAC address. But you forgot that the bitwise order of an _octet_ from a MAC address _memory_ and hexadecimal string representation is still always from left (MSB, bit 7) to right (LSB, bit 0). And the standard deals with memory representations only, so the transmission order of a MAC doesnt' matter here. As mentioned, OSSP uuid already implements this correctly. The FreeBSD uuidgen(2) and Apache APR generators I've also fixed myself recently in CVS. And for the remaining implementations I've submitted patches to the authors and they all (except for WINE) responded that they took over the patch. So the results of this long-standing bug we were able to fix -- at least for the free software world ;-). What is now remaining is that you finally also should fix this in your standard so the bug does not spread any longer into other implementations. Here is the minimal required patch against your draft: --- draft-mealling-uuid-urn-02.txt.orig Mon Feb 2 21:50:35 2004 +++ draft-mealling-uuid-urn-02.txt Fri Feb 13 15:41:49 2004 @@ -751,7 +751,7 @@ [6], and the cost was US$550. A better solution is to obtain a 47-bit cryptographic quality random - number, and use it as the low 47 bits of the node ID, with the most + number, and use it as the low 47 bits of the node ID, with the least significant bit of the first octet of the node ID set to one. This bit is the unicast/multicast bit, which will never be set in IEEE 802 addresses obtained from network cards; hence, there can never be a @@ -1369,7 +1369,7 @@ } else { get_random_info(seed); - seed[0] |= 0x80; + seed[0] |= 0x01; memcpy(&saved_node, seed, sizeof saved_node); fp = fopen("nodeid", "wb"); if (fp) { But I recommend you to perhaps also add one or two sentences which explain what I explained above (the difference between memory and transmission order), just to make sure people are not confused in the other direction and then think there is a bug (in the then fixed and correct) standard, because they know about the transmission order of MAC addresses. Yours, Ralf S. Engelschall rse@engelschall.com www.engelschall.com Date: Fri, 13 Feb 2004 11:05:51 -0500 From: Rich Salz To: rse@engelschall.com Cc: paulle@microsoft.com, michael@neonym.net Message-ID: <402CF5DF.4020601@datapower.com> Subject: Re: [PATCH] draft-mealling-uuid-urn-02.txt References: <20040213150931.GA7656@engelschall.com> Content-Length: 431 Lines: 11 Thanks for writing, Ralf. You're correct, and this has been noted by the IESG and will be fixed in the next draft. It's unfortunate we made it this far with the bug. /r$ -- Rich Salz, Chief Security Architect DataPower Technology http://www.datapower.com XS40 XML Security Gateway http://www.datapower.com/products/xs40.html XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html Date: Thu, 22 Jan 2004 05:34:11 -0800 (PST) From: "Ralf S. Engelschall" Message-Id: <200401221334.i0MDYB1K018137@repoman.freebsd.org> To: src-committers@FreeBSD.org, cvs-src@FreeBSD.org, cvs-all@FreeBSD.org Subject: cvs commit: src/sys/kern kern_uuid.c X-FreeBSD-CVS-Branch: HEAD X-Loop: FreeBSD.ORG Content-Length: 1907 Lines: 42 rse 2004/01/22 05:34:11 PST FreeBSD src repository Modified files: sys/kern kern_uuid.c Log: Fix generation of random multicast MAC address. In case no real/physical IEEE 802 address is available, both the expired "draft-leach-uuids-guids-01" (section "4. Node IDs when no IEEE 802 network card is available") and RFC 2518 (section "6.4.1 Node Field Generation Without the IEEE 802 Address") recommend (quoted from RFC 2518): "The ideal solution is to obtain a 47 bit cryptographic quality random number, and use it as the low 47 bits of the node ID, with the _most_ significant bit of the first octet of the node ID set to 1. This bit is the unicast/multicast bit, which will never be set in IEEE 802 addresses obtained from network cards; hence, there can never be a conflict between UUIDs generated by machines with and without network cards." Unfortunately, this incorrectly explains how to implement this and the FreeBSD UUID generator code inherited this generation bug from the broken reference code in the standards draft. They should instead specify the "_least_ significant bit of the first octet of the node ID" as the multicast bit in a memory and hexadecimal string representation of a 48-bit IEEE 802 MAC address. This standards bug arised from a false interpretation, as the multicast bit is actually the _most_ significant bit in IEEE 802.3 (Ethernet) _transmission order_ of an IEEE 802 MAC address. The standards authors forgot that the bitwise order of an _octet_ from a MAC address _memory_ and hexadecimal string representation is still always from left (MSB, bit 7) to right (LSB, bit 0). Fortunately, this UUID generation bug could have occurred on systems without any Ethernet NICs only. Revision Changes Path 1.7 +1 -1 src/sys/kern/kern_uuid.c Date: Thu, 22 Jan 2004 15:20:22 -0800 From: Marcel Moolenaar To: "Ralf S. Engelschall" Cc: src-committers@FreeBSD.org, cvs-src@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/kern kern_uuid.c Message-ID: <20040122232022.GA77798@ns1.xcllnt.net> References: <200401221334.i0MDYB1K018137@repoman.freebsd.org> Content-Length: 380 Lines: 14 On Thu, Jan 22, 2004 at 05:34:11AM -0800, Ralf S. Engelschall wrote: > rse 2004/01/22 05:34:11 PST > > FreeBSD src repository > > Modified files: > sys/kern kern_uuid.c > Log: > Fix generation of random multicast MAC address. An excellent catch and an outstanding commit log. Chapeau! -- Marcel Moolenaar USPA: A-39004 marcel@xcllnt.net ___________________________________________________________________________ Index: ChangeLog =================================================================== RCS file: /e/ossp/cvs/ossp-pkg/uuid/ChangeLog,v retrieving revision 1.42 diff -u -d -r1.42 ChangeLog --- ChangeLog 13 Feb 2004 16:17:07 -0000 1.42 +++ ChangeLog 13 Feb 2004 21:01:07 -0000 @@ -13,6 +13,14 @@ Changes between 0.9.6 and 0.9.7 (11-Feb-2004 to 13-Feb-2004) + o remove --with-rfc2518 option and functionality because + even the IETF/IESG has finally approved our report about the broken + random multicast MAC address generation in the standard (and + will fix it in new versions of the draft-mealling-uuid-urn). So, + finally get rid of this broken-by-design backward compatibility + functionality. + [Ralf S. Engelschall] + o Add support to uuid(1) CLI for decoding from stdin for both binary and string representations. [Ralf S. Engelschall] Index: uuid.ac =================================================================== RCS file: /e/ossp/cvs/ossp-pkg/uuid/uuid.ac,v retrieving revision 1.10 diff -u -d -r1.10 uuid.ac --- uuid.ac 11 Feb 2004 14:38:40 -0000 1.10 +++ uuid.ac 13 Feb 2004 19:20:32 -0000 @@ -71,12 +71,6 @@ AC_CHECK_SIZEOF(unsigned long long, 8) dnl # options - AC_ARG_WITH(rfc2518, - AC_HELP_STRING([--with-rfc2518], [use incorrect generation of IEEE 802 multicast addresses according to RFC2518]), - [ac_cv_with_rfc2518=$withval], [ac_cv_with_rfc2518=no]) - if test ".$ac_cv_with_rfc2518" = ".yes"; then - AC_DEFINE(WITH_RFC2518, 1, [whether to use incorrect generation of IEEE 802 multicast addresses according to RFC2518]) - fi AC_ARG_WITH(dce, AC_HELP_STRING([--with-dce], [build DCE 1.1 backward compatibility API]), [ac_cv_with_dce=$withval], [ac_cv_with_dce=no]) Index: uuid.c =================================================================== RCS file: /e/ossp/cvs/ossp-pkg/uuid/uuid.c,v retrieving revision 1.44 diff -u -d -r1.44 uuid.c --- uuid.c 19 Jan 2004 14:56:35 -0000 1.44 +++ uuid.c 13 Feb 2004 19:22:01 -0000 @@ -61,69 +61,9 @@ Unix UTC base time is January 1, 1970) */ #define UUID_TIMEOFFSET "01B21DD213814000" -/* IEEE 802 MAC address encoding/decoding bit fields - - ATTENTION: - - In case no real/physical IEEE 802 address is available, both - "draft-leach-uuids-guids-01" (section "4. Node IDs when no IEEE 802 - network card is available") and RFC 2518 (section "6.4.1 Node Field - Generation Without the IEEE 802 Address") recommend (quoted from RFC - 2518): - - "The ideal solution is to obtain a 47 bit cryptographic quality - random number, and use it as the low 47 bits of the node ID, with - the most significant bit of the first octet of the node ID set to - 1. This bit is the unicast/multicast bit, which will never be set - in IEEE 802 addresses obtained from network cards; hence, there can - never be a conflict between UUIDs generated by machines with and - without network cards." - - This passage clearly explains the intention to use IEEE 802 multicast - addresses. Unfortunately, it incorrectly explains how to implement - this! It should instead specify the "*LEAST* significant bit of the - first octet of the node ID" as the multicast bit in a memory and - hexadecimal string representation of a 48-bit IEEE 802 MAC address. - - Unfortunately, even the reference implementation included in the - expired IETF "draft-leach-uuids-guids-01" incorrectly set the - multicast bit with an OR bit operation and an incorrect mask of - 0x80. Hence, several other UUID implementations found on the - Internet have inherited this bug. - - Luckily, neither DCE 1.1 nor ISO/IEC 11578:1996 are affected by this - problem. They disregard the topic of missing IEEE 802 addresses - entirely, and thus avoid adopting this bug from the original draft - and code ;-) - - It seems that this standards bug arises from a false interpretation, - as the multicast bit is actually the *MOST* significant bit in IEEE - 802.3 (Ethernet) _transmission order_ of an IEEE 802 MAC address. The - authors were likely not aware that the bitwise order of an octet from - a MAC address memory and hexadecimal string representation is still - always from left (MSB, bit 7) to right (LSB, bit 0). - - For more information, see "Understanding Physical Addresses" in - "Ethernet -- The Definitive Guide", p.43, and the section "ETHERNET - MULTICAST ADDRESSES" in http://www.iana.org/assignments/ethernet-numbers. - - At OSSP, we do it the intended/correct way and generate a real - IEEE 802 multicast address. Those wanting to encode broken IEEE - 802 MAC addresses (as specified) can nevertheless use a brain dead - compile-time option to switch off the correct behavior. When decoding - we always use the correct behavior of course. */ - -/* encoding */ -#ifdef WITH_RFC2518 -#define IEEE_MAC_MCBIT_ENC BM_OCTET(1,0,0,0,0,0,0,0) -#else -#define IEEE_MAC_MCBIT_ENC BM_OCTET(0,0,0,0,0,0,0,1) -#endif -#define IEEE_MAC_LOBIT_ENC BM_OCTET(0,0,0,0,0,0,1,0) - -/* decoding */ -#define IEEE_MAC_MCBIT_DEC BM_OCTET(0,0,0,0,0,0,0,1) -#define IEEE_MAC_LOBIT_DEC BM_OCTET(0,0,0,0,0,0,1,0) +/* IEEE 802 MAC address encoding/decoding bit fields */ +#define IEEE_MAC_MCBIT BM_OCTET(0,0,0,0,0,0,0,1) +#define IEEE_MAC_LOBIT BM_OCTET(0,0,0,0,0,0,1,0) /* IEEE 802 MAC address octet length */ #define IEEE_MAC_OCTETS 6 @@ -622,8 +562,8 @@ (unsigned int)uuid->obj.node[3], (unsigned int)uuid->obj.node[4], (unsigned int)uuid->obj.node[5], - (uuid->obj.node[0] & IEEE_MAC_LOBIT_DEC ? "local" : "global"), - (uuid->obj.node[0] & IEEE_MAC_MCBIT_DEC ? "multicast" : "unicast")); + (uuid->obj.node[0] & IEEE_MAC_LOBIT ? "local" : "global"), + (uuid->obj.node[0] & IEEE_MAC_MCBIT ? "multicast" : "unicast")); } else { /* decode anything else as hexadecimal byte-string only */ @@ -843,8 +783,8 @@ if ((mode & UUID_MAKE_MC) || (uuid->mac[0] & BM_OCTET(1,0,0,0,0,0,0,0))) { /* generate random IEEE 802 local multicast MAC address */ prng_data(uuid->prng, (void *)&(uuid->obj.node), sizeof(uuid->obj.node)); - uuid->obj.node[0] |= IEEE_MAC_MCBIT_ENC; - uuid->obj.node[0] |= IEEE_MAC_LOBIT_ENC; + uuid->obj.node[0] |= IEEE_MAC_MCBIT; + uuid->obj.node[0] |= IEEE_MAC_LOBIT; } else { /* use real regular MAC address */