sr-dev November 2009

sr-dev@lists.kamailio.org

30 participants
177 discussions

git:master: tm: Number of fixes in code and documentation for serial forking.

by Jan Janak

Module: sip-router Branch: master Commit: ed3902e9bd5a3af6ee5d5df08ce2564f028a0f24 URL: http://git.sip-router.org/cgi-bin/gitweb.cgi/sip-router/?a=commit;h=ed3902e… Author: Jan Janak <jan(a)ryngle.com> Committer: Jan Janak <jan(a)ryngle.com> Date: Sun Oct 18 00:39:31 2009 +0200 tm: Number of fixes in code and documentation for serial forking. This patch improves the serial forking related code and documentation. The original code suffers from shortcommits, some of which were result of the migration to the new core. The function that decodes destination sets encoded in AVPs has been improved, because the original version did not handle properly strings with missing elements. In such case the original implementation was likely to overwrite memory, because it did not check the return value of strchr properly. The new implementation tries to handle this situation. It continues parsing as long as it can, it only requires that the request-uri string is present, all other fields are made optional and their variables are properly initialized if their values cannot be found in the AVP. There was a bug in the implementation of fr_inv_timer_next modparam in the original version, changes to the parameter value were ignored by the serial forking code. This was reported by Andrei and is now fixed by this commit. The parameter fr_inv_timer_next can now be configured at runtime with the configuration framework. Its value is in milliseconds and unlike fr_inv_timer, this timer cannot be configured separately for individual branches. Obsolete definition of INV_FR_TIME_OUT_FIRST has been removed. That macro is not used anywhere in the code, thus it is not needed. There were several places where LM_DBG printed a string and relied on the string being zero terminated, this may or may not be true in the future and this patch uses the macro STR_FMT where appropriate, which is safer. Function t_next_contacts now checks if the function decode_branch_info really returned values for the dst_uri and the path vector. If not then it calls reset_dst_uri and reset_path_vector. Previous version of the code crashed sip-router, this is likely due to the merge and updates in the sip-router core. We now use t_set_fr in t_next_contacts for setting the fr_inv_timer value to fr_inv_timer_next. This is much more efficient than creating AVPs with new timer values. Also the new value of the timer is now taken from a variable in the configuration framework, instead of just a regular global variable configured through modparam. This way we can adjust the value of the timer on the fly. Configuring it through modparam is, of course, possible too. The value of of fr_inv_timer_next is now in milliseconds, instead of seconds. That's the only possibly incompatible change. However, this is consistent with all other timers in tm module, it is more efficient and it offers better granularity. A missing call to destroy_avp has been added to t_next_contacts, in the code which is executed when no transaction exists. There, the avp should also be destroyed if all values have been exhausted and none of them had Q_FLAG set. This is a corner case which should not happen under normal circumstances, because that situation only happens if all branches have the same q value. Such AVP would not have been created by t_load_contacts and therefore t_next_contacts should not be called, but this bug may be triggered if someone uses t_next_contacts in an unexpected way and it is probably better to have it fixed. Also the code which restores the value of fr_inv_timer at the end of t_next_contacts did not work properly. This patch fixes that. It first tries to retrieve a value configured with t_set_fr, but that is not guaranteed to succeed. After that it also tries the timer AVP and finally the configuration framework. The configuration framework always yields a value, so we can always restore the timer value, but we may fail to restore individual transaction timer values set by t_set_fr. If that fails then the global value from the configuration framework is used. This is documented as a shortcomming in the README and in the code. In addition to code changes this patch also expands documentation on functions t_load_contacts and t_next_contacts, describing their operation in more detail. Also the format of the contacts AVP is now documented. Finally, there is a whole new section in the README which describes how serial/parallel forking can be achieved with t_load_contacts and t_next_contacts and provides a number of examples. --- modules/tm/README | 592 +++++++++++++++++++++++++++--------------- modules/tm/config.c | 7 +- modules/tm/config.h | 3 +- modules/tm/doc/functions.xml | 129 ++++++++-- modules/tm/doc/params.xml | 83 +++++-- modules/tm/doc/tm.xml | 218 ++++++++++++++++ modules/tm/t_serial.c | 208 ++++++++++------ modules/tm/tm.c | 2 +- 8 files changed, 900 insertions(+), 342 deletions(-) Diff: http://git.sip-router.org/cgi-bin/gitweb.cgi/sip-router/?a=commitdiff;h=ed3…

15 years

stored message are not removed after delivery

by jh＠tutpro.com

today i noticed that stored messages are not removed from silo table after delivery and they will be delivered again target ua re-registers after an hour. target ua receives the stored messages normally and wireshark shows that 200 ok gets back to sr fine. sr does not try to resend message, so i assume that its transaction layer accept the 200 ok. any ideas why delivered messages stay in silo table. this has worked fine in k. -- juha

15 years

SF.net SVN: openser:[5952] branches/1.5/modules/sanity/sanity.c

by Daniel-Constantin Mierla

Revision: 5952 http://openser.svn.sourceforge.net/openser/?rev=5952&view=rev Author: miconda Date: 2009-11-01 15:56:46 +0000 (Sun, 01 Nov 2009) Log Message: ----------- - one more proper return check for r-uri parsing Modified Paths: -------------- branches/1.5/modules/sanity/sanity.c This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.

15 years

SF.net SVN: openser:[5951] branches/1.5/modules/pv/pv_trans.c

by Daniel-Constantin Mierla

Revision: 5951 http://openser.svn.sourceforge.net/openser/?rev=5951&view=rev Author: miconda Date: 2009-11-01 15:50:10 +0000 (Sun, 01 Nov 2009) Log Message: ----------- - reset entire local structure holding To value instead of just len Modified Paths: -------------- branches/1.5/modules/pv/pv_trans.c This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.

15 years

$ai does not work

by Juha Heinanen

i started to wonder why some stuff is missing from accounting records and looks like the reason is that pv $ai is not working. i have in my script: if (is_present_hf("P-Asserted-Identity")) { xlog("L_INFO", "P-A-I header <$(hdr(P-Asserted-Identity))> with URI <$ai>\n"); and i get to syslog: Nov 1 10:50:19 localhost /usr/sbin/sip-proxy[1241]: INFO: P-A-I header <sip:jh@test.fi> with URI <<null>> -- juha

15 years

git:sr_3.0: core: PAI and PPI parsing support fixed

by Juha Heinanen

Module: sip-router Branch: sr_3.0 Commit: 2163a7f7706a70c44faef18d08a89dbc960bfd47 URL: http://git.sip-router.org/cgi-bin/gitweb.cgi/sip-router/?a=commit;h=2163a7f… Author: Daniel-Constantin Mierla <miconda(a)gmail.com> Committer: Juha Heinanen <jh(a)tutpro.com> Date: Sun Nov 1 14:06:04 2009 +0100 core: PAI and PPI parsing support fixed - use a new function to identify hdr type instead of 'case' construct - reported by Juha Heinanen (cherry picked from commit 78f069c3726085902101a55d3bb98f0be7549b60) --- parser/hf.c | 9 +++++++++ parser/msg_parser.c | 10 ++++++++++ parser/parse_hname2.c | 20 ++++++++++++++++++++ 3 files changed, 39 insertions(+), 0 deletions(-) diff --git a/parser/hf.c b/parser/hf.c index a54865e..09fcc4a 100644 --- a/parser/hf.c +++ b/parser/hf.c @@ -216,6 +216,15 @@ void clean_hdr_field(struct hdr_field* hf) case HDR_PATH_T: case HDR_PRIVACY_T: break; + + case HDR_PPI_T: + free_to(hf->parsed); + break; + + case HDR_PAI_T: + free_to(hf->parsed); + break; + default: LOG(L_CRIT, "BUG: clean_hdr_field: unknown header type %d\n", hf->type); diff --git a/parser/msg_parser.c b/parser/msg_parser.c index 0388bd4..a3531e2 100644 --- a/parser/msg_parser.c +++ b/parser/msg_parser.c @@ -243,6 +243,8 @@ char* get_hdr_field(char* buf, char* end, struct hdr_field* hdr) case HDR_PROXY_AUTHENTICATE_T: case HDR_PATH_T: case HDR_PRIVACY_T: + case HDR_PAI_T: + case HDR_PPI_T: case HDR_OTHER_T: /* just skip over it */ hdr->body.s=tmp; @@ -528,6 +530,14 @@ int parse_headers(struct sip_msg* msg, hdr_flags_t flags, int next) if (msg->privacy==0) msg->privacy=hf; msg->parsed_flag|=HDR_PRIVACY_F; break; + case HDR_PAI_T: + if (msg->pai==0) msg->pai=hf; + msg->parsed_flag|=HDR_PAI_F; + break; + case HDR_PPI_T: + if (msg->ppi==0) msg->ppi=hf; + msg->parsed_flag|=HDR_PPI_F; + break; default: LOG(L_CRIT, "BUG: parse_headers: unknown header type %d\n", hf->type); diff --git a/parser/parse_hname2.c b/parser/parse_hname2.c index 2b4152a..f94a30d 100644 --- a/parser/parse_hname2.c +++ b/parser/parse_hname2.c @@ -163,6 +163,24 @@ static inline char* skip_ws(char* p, unsigned int size) } +int hdr_update_type(struct hdr_field* hdr) +{ + if(hdr==0 || hdr->name.s==0) + return -1; + switch(hdr->name.len) { + case 19: + if(strncasecmp(hdr->name.s, "P-Asserted-Identity", 19)==0) + hdr->type = HDR_PAI_T; + break; + case 20: + if(strncasecmp(hdr->name.s, "P-Preferred-Identity", 20)==0) + hdr->type = HDR_PPI_T; + break; + } + + return 0; +} + char* parse_hname2(char* begin, char* end, struct hdr_field* hdr) { register char* p; @@ -242,6 +260,8 @@ char* parse_hname2(char* begin, char* end, struct hdr_field* hdr) } else { hdr->type = HDR_OTHER_T; hdr->name.len = p - hdr->name.s; + hdr_update_type(hdr); return (p + 1); } } +

15 years

git:kamailio_3.0: core: PAI and PPI parsing support fixed

by Daniel-Constantin Mierla

Module: sip-router Branch: kamailio_3.0 Commit: 78f069c3726085902101a55d3bb98f0be7549b60 URL: http://git.sip-router.org/cgi-bin/gitweb.cgi/sip-router/?a=commit;h=78f069c… Author: Daniel-Constantin Mierla <miconda(a)gmail.com> Committer: Daniel-Constantin Mierla <miconda(a)gmail.com> Date: Sun Nov 1 14:06:04 2009 +0100 core: PAI and PPI parsing support fixed - use a new function to identify hdr type instead of 'case' construct - reported by Juha Heinanen --- parser/hf.c | 9 +++++++++ parser/msg_parser.c | 10 ++++++++++ parser/parse_hname2.c | 20 ++++++++++++++++++++ 3 files changed, 39 insertions(+), 0 deletions(-) diff --git a/parser/hf.c b/parser/hf.c index a54865e..09fcc4a 100644 --- a/parser/hf.c +++ b/parser/hf.c @@ -216,6 +216,15 @@ void clean_hdr_field(struct hdr_field* hf) case HDR_PATH_T: case HDR_PRIVACY_T: break; + + case HDR_PPI_T: + free_to(hf->parsed); + break; + + case HDR_PAI_T: + free_to(hf->parsed); + break; + default: LOG(L_CRIT, "BUG: clean_hdr_field: unknown header type %d\n", hf->type); diff --git a/parser/msg_parser.c b/parser/msg_parser.c index d218b0b..e5cb7e6 100644 --- a/parser/msg_parser.c +++ b/parser/msg_parser.c @@ -244,6 +244,8 @@ char* get_hdr_field(char* buf, char* end, struct hdr_field* hdr) case HDR_PROXY_AUTHENTICATE_T: case HDR_PATH_T: case HDR_PRIVACY_T: + case HDR_PAI_T: + case HDR_PPI_T: case HDR_OTHER_T: /* just skip over it */ hdr->body.s=tmp; @@ -530,6 +532,14 @@ int parse_headers(struct sip_msg* msg, hdr_flags_t flags, int next) if (msg->privacy==0) msg->privacy=hf; msg->parsed_flag|=HDR_PRIVACY_F; break; + case HDR_PAI_T: + if (msg->pai==0) msg->pai=hf; + msg->parsed_flag|=HDR_PAI_F; + break; + case HDR_PPI_T: + if (msg->ppi==0) msg->ppi=hf; + msg->parsed_flag|=HDR_PPI_F; + break; default: LOG(L_CRIT, "BUG: parse_headers: unknown header type %d\n", hf->type); diff --git a/parser/parse_hname2.c b/parser/parse_hname2.c index 2b4152a..f94a30d 100644 --- a/parser/parse_hname2.c +++ b/parser/parse_hname2.c @@ -163,6 +163,24 @@ static inline char* skip_ws(char* p, unsigned int size) } +int hdr_update_type(struct hdr_field* hdr) +{ + if(hdr==0 || hdr->name.s==0) + return -1; + switch(hdr->name.len) { + case 19: + if(strncasecmp(hdr->name.s, "P-Asserted-Identity", 19)==0) + hdr->type = HDR_PAI_T; + break; + case 20: + if(strncasecmp(hdr->name.s, "P-Preferred-Identity", 20)==0) + hdr->type = HDR_PPI_T; + break; + } + + return 0; +} + char* parse_hname2(char* begin, char* end, struct hdr_field* hdr) { register char* p; @@ -242,6 +260,8 @@ char* parse_hname2(char* begin, char* end, struct hdr_field* hdr) } else { hdr->type = HDR_OTHER_T; hdr->name.len = p - hdr->name.s; + hdr_update_type(hdr); return (p + 1); } } +

15 years

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

sr-dev November 2009