openldap-bugs

openldap-bugs@openldap.org

1 participants
20963 discussions

(ITS#8235) compiler warnings in smbk5pwd.c with nettle
by ryan＠openldap.org 02 Sep '15

02 Sep '15

Full_Name: Ryan Tandy Version: master OS: Debian URL: Submission from: (NULL) (24.68.37.4) Submitted by: ryan smbk5pwd.c: In function 'lmhash': smbk5pwd.c:206:21: warning: passing argument 2 of 'nettle_des_set_key' from incompatible pointer type [-Wincompatible-pointer-types] des_set_key( &ctx, &key ); ^AIAIn file included from smbk5pwd.c:69:0: /usr/include/nettle/des.h:76:1: note: expected 'const uint8_t * {aka const unsigned char *}' but argument is of type 'unsigned char (*)[8]' des_set_key(struct des_ctx *ctx, const uint8_t *key); ^ (and several more) The openssl and gcrypt versions both used "&key" as well. It's correct for openssl, which takes a DES_cblock *. It's ambiguous for gcrypt, which takes a void *. It's wrong for nettle, which expects an unsigned char *. This is not a functional problem, since in all cases the pointer points at the beginning of the array. Fixing the warning does not change the output. smbk5pwd.c: In function 'nthash': smbk5pwd.c:249:36: warning: pointer targets in passing argument 3 of 'nettle_md4_update' differ in signedness [-Wpointer-sign] md4_update( &ctx, passwd->bv_len, passwd->bv_val ); ^ In file included from smbk5pwd.c:70:0: /usr/include/nettle/md4.h:69:1: note: expected 'const uint8_t * {aka const unsigned char *}' but argument is of type 'char *' md4_update(struct md4_ctx *ctx, ^ openssl and gcrypt both use void * here, nettle needs a cast (which should be safe). The remaining warnings are not nettle-specific. smbk5pwd.c: In function 'smbk5pwd_cf_func': smbk5pwd.c:915:14: warning: variable 'db' set but not used [-Wunused-but-set-variable] BackendDB db = *c->be; ^ Added in 29fab9c, apparently never used. Looks like cruft to me. Most likely didn't generate a warning with gcc prior to 4.6, due to the assignment just below. smbk5pwd.c: In function 'smbk5pwd_modules_init': smbk5pwd.c:968:2: warning: variable 'dummy_ad' set but not used [-Wunused-but-set-variable] dummy_ad; ^ The unused-but-set-variable warning is new in gcc 4.6 and there is a different no-op trick to suppress it in a portable way.

1 0

(ITS#8234) ppolicy_get can return a partly-populated struct
by ryan＠openldap.org 01 Sep '15

01 Sep '15

Full_Name: Ryan Tandy Version: master OS: URL: Submission from: (NULL) (24.68.37.4) Submitted by: ryan In ppolicy_get, if parsing fails after some attributes have already been loaded, we can return a partly populated structure, while the warning says "using default policy". In this case we should revert the entire structure to defaults before returning, IMO.

1 0

Re: (ITS#7537) ppolicy hangs slapd on 64bit version debian
by ryan＠nardis.ca 01 Sep '15

01 Sep '15

On Tue, Aug 25, 2015 at 10:00:55PM +0000, ryan(a)nardis.ca wrote: >I reproduced this with current RE24. It seems to be caused by setting >pwdMaxAge to a very large value. ... or any other ppolicy attribute. >It seems like the crash does not happen if I bind as the rootdn, only >when using the admin DN. That was incorrect, bind DN doesn't matter. In ppolicy_get, the defaultpol label skips releasing the entry lock, so if it's called again, it deadlocks.

1 0

(ITS#8233) config parser accepts lines with unbalanced quotes
by hyc＠openldap.org 31 Aug '15

31 Aug '15

Full_Name: Howard Chu Version: 2.4 OS: URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (50.232.101.67) Submitted by: hyc We were tripped up by a misbehaving syncrepl consumer, that was always connecting as plain syncrepl even though it was intended to use delta-sync. The config had been pasted out of an email and the mail client had turned some (but not all) ASCII double-quotes into UTF-8 smart quotes. As such, the syncrepl stanza ended at logfilter="xxxxx and this wasn't flagged as an error because 1) the config parser doesn't care if double quotes get terminated or not 2) the syncrepl config parser doesn't look at logfilter if logbase wasn't configured. The logbase token was swallowed up in the logfilter. The parser should be fixed to reject lines with unbalanced quotes.

1 0

Re: (ITS#8230) [PATCH] totp: bug fixes and improvements
by hyc＠symas.com 31 Aug '15

31 Aug '15

Peter Marschall wrote: > Hi, > > On Sunday, 30. August 2015 23:58:09 Howard Chu wrote: >>> Please note the run index i changed direction, which causes different >>> results because of >>=. >>> >>> In my tests [double checked before sending this mail] on an amd64 [little- >>> endian architecture], the patched version converts 0x3837363534333231 to >>> - "87654321" in the !WORDS_BIGENDIAN case [unchanged] >>> - "12345678" in the WORDS_BIGENDIAN case >>> [using the fact that 0x3N = "N" for 0 <= N <=9] >>> To me that looks correct. >>> >>> What issues did you see exactly? >> >> That is exactly wrong. Given any integer input the string should be >> identical regardless of endianness. It should be "87654321" in both cases. >> With your patch, hashes will be incompatible between different machine >> architectures. > > Please read my mail again! > I tested the WORDS_BIGENDIAN case on a *little-endian* architecture using the > same data as the !WORDS_BIGENDIAN case. > > In that case [i.e. testing on the "wrong-endian" architecture] the result *is > expected* to come out in reversed order. > Otherwise it would end-up in wrong order on the "right-endian" architecture. > > Do you agree? The point is that your patch *always* reverses the order of the bytes. On a big-endian machine the code should be a no-op. > > If you don't, please hint me to some code that allows me to check and fix my > code. > > Best > Peter > -- -- Howard Chu CTO, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc/ Chief Architect, OpenLDAP http://www.openldap.org/project/

1 0

Re: (ITS#8230) [PATCH] totp: bug fixes and improvements
by peter＠adpm.de 31 Aug '15

31 Aug '15

Hi, On Sunday, 30. August 2015 23:58:09 Howard Chu wrote: > > Please note the run index i changed direction, which causes different > > results because of >>=. > > > > In my tests [double checked before sending this mail] on an amd64 [little- > > endian architecture], the patched version converts 0x3837363534333231 to > > - "87654321" in the !WORDS_BIGENDIAN case [unchanged] > > - "12345678" in the WORDS_BIGENDIAN case > > [using the fact that 0x3N = "N" for 0 <= N <=9] > > To me that looks correct. > > > > What issues did you see exactly? > > That is exactly wrong. Given any integer input the string should be > identical regardless of endianness. It should be "87654321" in both cases. > With your patch, hashes will be incompatible between different machine > architectures. Please read my mail again! I tested the WORDS_BIGENDIAN case on a *little-endian* architecture using the same data as the !WORDS_BIGENDIAN case. In that case [i.e. testing on the "wrong-endian" architecture] the result *is expected* to come out in reversed order. Otherwise it would end-up in wrong order on the "right-endian" architecture. Do you agree? If you don't, please hint me to some code that allows me to check and fix my code. Best Peter -- Peter Marschall peter(a)adpm.de

1 0

Re: (ITS#8230) [PATCH] totp: bug fixes and improvements
by hyc＠symas.com 30 Aug '15

30 Aug '15

Peter Marschall wrote: >>> For the big-endian case, 'msg' wasn't set from 'tval' in generate(). >> >> This patch is wrong, as I commented on github. > Are you sure? > > Please note the run index i changed direction, which causes different results > because of >>=. > > In my tests [double checked before sending this mail] on an amd64 [little- > endian architecture], the patched version converts 0x3837363534333231 to > - "87654321" in the !WORDS_BIGENDIAN case [unchanged] > - "12345678" in the WORDS_BIGENDIAN case > [using the fact that 0x3N = "N" for 0 <= N <=9] > To me that looks correct. > > What issues did you see exactly? That is exactly wrong. Given any integer input the string should be identical regardless of endianness. It should be "87654321" in both cases. With your patch, hashes will be incompatible between different machine architectures. -- -- Howard Chu CTO, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc/ Chief Architect, OpenLDAP http://www.openldap.org/project/

1 0

Re: (ITS#8230) [PATCH] totp: bug fixes and improvements
by peter＠adpm.de 30 Aug '15

30 Aug '15

Hi Howard, thanks for the quick response. On Saturday, 29. August 2015 16:43:59 Howard Chu wrote: > > In function totp_b32_pton() > > - allow lowercase characters in encoded string too > > This patch is incorrect. You must guard toupper(), it is not guaranteed not > to distort non-lowercase characters. > > if (islower(x)) x = toupper(x) Thx, will include it in a reworked version. > > - allow padding to be omitted (totally, not only parts) > > Why? To allow using the keys encoded by other implementations that do not generate the padding (e.g. Perl's Convert::Base32). (e.g. in a mass-rollout that sets userPassword using LDIF) > > For the big-endian case, 'msg' wasn't set from 'tval' in generate(). > > This patch is wrong, as I commented on github. Are you sure? Please note the run index i changed direction, which causes different results because of >>=. In my tests [double checked before sending this mail] on an amd64 [little- endian architecture], the patched version converts 0x3837363534333231 to - "87654321" in the !WORDS_BIGENDIAN case [unchanged] - "12345678" in the WORDS_BIGENDIAN case [using the fact that 0x3N = "N" for 0 <= N <=9] To me that looks correct. What issues did you see exactly? > > In totp_b32_pton(), correctly count the number of '=' padding chars > > at the end of the base-32 encoded string. > > > > Note: '*str++' evaluates *str first and increases str later! > > The current code correctly decodes known test data. What is your test case? Hmm, strange. On my 64-bit Linux the original implementation failed to decode keys encoded by the original implementation's encoder. It took me quite some time to find the exact cause. I tested my patch with the official test vectors of chapter 10 of RFC 4648. With my patch the decoding in totp_b32_pton() is symmetric to encoding performed by totp_b32_ntop(), i.e. encoded strings decode to the original. In addition testing reproduced the official RFC4648 test cases in both directions, where the original version miscalculated the number of padding bytes so that "state + i" did not correctly add up to 8 as required for the "good" cases. I double checked before sending this mail: The original implementation fails to decode all test cases that have padding. E.g. BASE32("foo") = "MZXW6===" [RFC4648] In this example it counts the number of passing chars "=" to i = 4, which I do not consider correct. Would you mind posting your test cases? BTW: not insisting on padding would help here too ;-) > > * > > https://github.com/marschap/openldap/commit/04c15b7b1e44d4d3167577702a8b7 > > 7c5ee8bb7b6> > > contrib/passwd/totp: support compiling using nettle > > The proper way to add this support is to add macros to hide all of the > differences in the initial ifdef block and not to add any additional ifdefs > anywhere in the main body of the code. I'll have a closer look. This patch tried to not touch the original OpenSSL version. With a more "homogenous patch" I'll need to make some slight changes to the macros for the OpenSSL version. I hope that's OK. Best PEter -- Peter Marschall peter(a)adpm.de

1 0

Re: (ITS#8230) [PATCH] totp: bug fixes and improvements
by h.b.furuseth＠usit.uio.no 29 Aug '15

29 Aug '15

On 29/08/15 17:44, hyc(a)symas.com wrote: > This patch is incorrect. You must guard toupper(), it is not guaranteed not to > distort non-lowercase characters. Yes it is - for values in the range of unsigned char and EOF. Behavior is undefined for other values. The bug is failing to cast a char arguments to <ctype.h> functions to unsigned char. It's a common bug, so some implementations support char args. They cannot support that fully for (char)EOF, though. > if (islower(x)) x = toupper(x) That might hide the bug at times, maybe that's why you use it. But it's not reliable, islower too expects EOF or unsigned char. Use x = toupper((unsigned char)x).

1 0

(ITS#8232) potential crash from syncprov_op_abandon
by hyc＠openldap.org 29 Aug '15

29 Aug '15

Full_Name: Howard Chu Version: 2.4 OS: URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (73.15.219.210) Submitted by: hyc If a connection closes while syncprov_search_response is turning the current search op into a detached psearch op, connection_abandon() may get run on the connection while both the original and the detached copy of the op are present on the connection queue. detach_op will append the new copy onto the end of the connection's queue. connection_abandon will find the original op first, and call into syncprov_op_abandon. syncprov_op_abandon will cause the detached copy to be freed from the si->si_ops list (since both copies have same connid and msgid). If there has been no other activity on the connection, then the original op's o_next will be pointing to the detached copy. connection_abandon is using this o_next to iterate thru the connection's queue. After syncprov frees this copy, connection_abandon will probably SEGV. The fix is to prevent connection_abandon from calling abandon handlers on an op that has already been abandoned.

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

openldap-bugs