From: "Eli Dorfman (Voltaire)" <dorfman.eli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
To: Slava Strebkov <slavas-smomgflXvOZWk0Htik3J/w@public.gmane.org>
Cc: kliteyn-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org,
linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: Re: [PATCH v3] opensm: support routing engine update
Date: Sun, 14 Feb 2010 12:18:00 +0200 [thread overview]
Message-ID: <4B77CDD8.2040900@gmail.com> (raw)
In-Reply-To: <39C75744D164D948A170E9792AF8E7CA01F6FA30-QfUkFaTmzUSUvQqKE/ONIwC/G2K4zDHf@public.gmane.org>
Slava Strebkov wrote:
> HI!
> That was my misunderstanding - upon heavy sweep SM will try to load
> routing engines as defined in the SM.conf file (ftree & updn - in that
> order).
> So ftree will be loaded when switch comes back from reboot.
>
> Slava
>
> -----Original Message-----
> From: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> [mailto:linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org] On Behalf Of Yevgeny Kliteynik
> Sent: Monday, December 07, 2009 10:22 AM
> To: Slava Strebkov
> Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> Subject: Re: [PATCH v3] opensm: support routing engine update
>
> Slava,
>
> Slava Strebkov wrote:
>> Hi Yevgeny,
>> In that case SM will use updn and will not come back to ftree
>> automatically.
>
> I think that this is a bad thing.
>
> I wouldn't want *temporary* change of fabric to cause
> *permanent* change of SM mode of operation. Such changes
> do happen, and I'd prefer SM to continue functioning
> in accordance to the user's configuration once the
> fabric is settled again.
>
> I do see the cases were the change that you propose is
> beneficial - if fabric topology doesn't fits the chosen
> routing, SM will waste time on retrying the wrong routing
> at every heavy sweep, but this happens due to suboptimal
> SM configuration and not as a result of some event that
> user has no control of.
For every heavy sweep the SM will try to configure the routing engines as specified in the conf file.
So when a switch goes up and ftree configuration is valid, the SM will configure ftree instead of updn.
Eli
>
> -- Yevgeny
>
>
>
>> Slava
>>
>> -----Original Message-----
>> From: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
>> [mailto:linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org] On Behalf Of Yevgeny
> Kliteynik
>> Sent: Sunday, December 06, 2009 6:03 PM
>> To: Slava Strebkov
>> Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
>> Subject: Re: [PATCH v3] opensm: support routing engine update
>>
>> Slava,
>>
>> Slava Strebkov wrote:
>>> setup routing engine when in use and delete when failed.
>>> setup routing engine before use.
>>> delete resources when routing algorithm fails.
>>> this will save allocation for routing algorithms that are not used.
>> Suppose a user runs SM with ftree & updn routings (in that order),
>> and SM manages to route the fabric with ftree. At some point
>> some switch reboots and causes ftree to fail and SM routes the
>> fabric with updn.
>> Does this mean that ftree will be removed from the list, and
>> when the switch comes back, SM won't try ftree any more?
>>
>> -- Yevgeny
>>
>>> Signed-off-by: Slava Strebkov <slavas-smomgflXvOZWk0Htik3J/w@public.gmane.org>
>>> ---
>>> opensm/include/opensm/osm_opensm.h | 5 +++
>>> opensm/opensm/osm_opensm.c | 57
>> +++++++++++++++++++++++++++++++-----
>>> opensm/opensm/osm_subnet.c | 7 ++++-
>>> opensm/opensm/osm_ucast_mgr.c | 28 +++++++++++++++++
>>> 4 files changed, 88 insertions(+), 9 deletions(-)
>>>
>>> diff --git a/opensm/include/opensm/osm_opensm.h
>> b/opensm/include/opensm/osm_opensm.h
>>> index c121be4..ca0fddb 100644
>>> --- a/opensm/include/opensm/osm_opensm.h
>>> +++ b/opensm/include/opensm/osm_opensm.h
>>> @@ -109,6 +109,7 @@ typedef enum _osm_routing_engine_type {
>>> } osm_routing_engine_type_t;
>>> /***********/
>>>
>>> +struct osm_opensm;
>>> /****s* OpenSM: OpenSM/osm_routing_engine
>>> * NAME
>>> * struct osm_routing_engine
>>> @@ -122,6 +123,8 @@ typedef enum _osm_routing_engine_type {
>>> struct osm_routing_engine {
>>> const char *name;
>>> void *context;
>>> + int initialized;
>>> + int (*setup) (struct osm_routing_engine *re, struct osm_opensm
>> *p_osm);
>>> int (*build_lid_matrices) (void *context);
>>> int (*ucast_build_fwd_tables) (void *context);
>>> void (*ucast_dump_tables) (void *context);
>>> @@ -183,6 +186,7 @@ typedef struct osm_opensm {
>>> cl_dispatcher_t disp;
>>> cl_plock_t lock;
>>> struct osm_routing_engine *routing_engine_list;
>>> + struct osm_routing_engine *last_routing_engine;
>>> osm_routing_engine_type_t routing_engine_used;
>>> osm_stats_t stats;
>>> osm_console_t console;
>>> @@ -522,6 +526,7 @@ extern volatile unsigned int osm_exit_flag;
>>> * DESCRIPTION
>>> * Set to one to cause all threads to leave
>>> *********/
>>> +void osm_update_routing_engines(osm_opensm_t *osm, const char
>> *engine_names);
>>>
>>> END_C_DECLS
>>> #endif /* _OSM_OPENSM_H_ */
>>> diff --git a/opensm/opensm/osm_opensm.c b/opensm/opensm/osm_opensm.c
>>> index 50d1349..f90584d 100644
>>> --- a/opensm/opensm/osm_opensm.c
>>> +++ b/opensm/opensm/osm_opensm.c
>>> @@ -169,14 +169,7 @@ static void setup_routing_engine(osm_opensm_t
>> *osm, const char *name)
>>> memset(re, 0, sizeof(struct
>> osm_routing_engine));
>>>
>>> re->name = m->name;
>>> - if (m->setup(re, osm)) {
>>> - OSM_LOG(&osm->log, OSM_LOG_VERBOSE,
>>> - "setup of routing"
>>> - " engine \'%s\' failed\n",
>> name);
>>> - return;
>>> - }
>>> - OSM_LOG(&osm->log, OSM_LOG_DEBUG,
>>> - "\'%s\' routing engine set up\n",
>> re->name);
>>> + re->setup = m->setup;
>>> append_routing_engine(osm, re);
>>> return;
>>> }
>>> @@ -236,6 +229,54 @@ static void destroy_routing_engines(osm_opensm_t
>> *osm)
>>> r->delete(r->context);
>>> free(r);
>>> }
>>> + osm->routing_engine_list = NULL;
>>> +}
>>> +
>>> +static void update_routing_engine(
>>> + struct osm_routing_engine *cur,
>>> + struct osm_routing_engine *last)
>>> +{
>>> + struct osm_routing_engine *next = cur->next;
>>> + if (!last)
>>> + return; /* no last routing engine */
>>> + memcpy(cur, last, sizeof(*cur));
>>> + /* restore next */
>>> + cur->next = next;
>>> +}
>>> +
>>> +void osm_update_routing_engines(osm_opensm_t *osm, const char
>> *engine_names)
>>> +{
>>> + struct osm_routing_engine *r, *l;
>>> + /* find used routing engine and save as last */
>>> + l = r = osm->routing_engine_list;
>>> + if (r && osm->routing_engine_used ==
>> osm_routing_engine_type(r->name)) {
>>> + osm->last_routing_engine = r;
>>> + osm->routing_engine_list = r->next;
>>> + }
>>> + else while ((r = r->next)) {
>>> + if (osm->routing_engine_used ==
>>> + osm_routing_engine_type(r->name)) {
>>> + osm->last_routing_engine = r;
>>> + l->next = r->next;
>>> + break;
>>> + }
>>> + l = r;
>>> + }
>>> + /* cleanup prev routing engine list and replace with current
>> list */
>>> + destroy_routing_engines(osm);
>>> + setup_routing_engines(osm, engine_names);
>>> + /* check if last routing engine exist in new list and update
>> callbacks */
>>> + r = osm->routing_engine_list;
>>> + while (r) {
>>> + if (osm->routing_engine_used ==
>>> + osm_routing_engine_type(r->name)) {
>>> + update_routing_engine(r,
>> osm->last_routing_engine);
>>> + free(osm->last_routing_engine);
>>> + osm->last_routing_engine = NULL;
>>> + break;
>>> + }
>>> + r = r->next;
>>> + }
>>> }
>>>
>>>
> /**********************************************************************
>>> diff --git a/opensm/opensm/osm_subnet.c b/opensm/opensm/osm_subnet.c
>>> index 8d63a75..742ae64 100644
>>> --- a/opensm/opensm/osm_subnet.c
>>> +++ b/opensm/opensm/osm_subnet.c
>>> @@ -152,6 +152,11 @@ static void opts_setup_sm_priority(osm_subn_t
>> *p_subn, void *p_val)
>>> osm_set_sm_priority(p_sm, sm_priority);
>>> }
>>>
>>> +static void opts_setup_routing_engine(osm_subn_t *p_subn, void
>> *p_val)
>>> +{
>>> + osm_update_routing_engines(p_subn->p_osm, p_val);
>>> +}
>>> +
>>> static void opts_parse_net64(IN osm_subn_t *p_subn, IN char *p_key,
>>> IN char *p_val_str, void *p_v1, void *p_v2,
>>> void (*pfn)(osm_subn_t *, void *))
>>> @@ -324,7 +329,7 @@ static const opt_rec_t opt_tbl[] = {
>>> { "hop_weights_file", OPT_OFFSET(hop_weights_file),
>> opts_parse_charp, NULL, 0 },
>>> { "port_profile_switch_nodes",
>> OPT_OFFSET(port_profile_switch_nodes), opts_parse_boolean, NULL, 1 },
>>> { "sweep_on_trap", OPT_OFFSET(sweep_on_trap),
>> opts_parse_boolean, NULL, 1 },
>>> - { "routing_engine", OPT_OFFSET(routing_engine_names),
>> opts_parse_charp, NULL, 0 },
>>> + { "routing_engine", OPT_OFFSET(routing_engine_names),
>> opts_parse_charp, opts_setup_routing_engine, 1 },
>>> { "connect_roots", OPT_OFFSET(connect_roots),
>> opts_parse_boolean, NULL, 1 },
>>> { "use_ucast_cache", OPT_OFFSET(use_ucast_cache),
>> opts_parse_boolean, NULL, 1 },
>>> { "log_file", OPT_OFFSET(log_file), opts_parse_charp, NULL, 0 },
>>> diff --git a/opensm/opensm/osm_ucast_mgr.c
>> b/opensm/opensm/osm_ucast_mgr.c
>>> index 39d825c..d6294ac 100644
>>> --- a/opensm/opensm/osm_ucast_mgr.c
>>> +++ b/opensm/opensm/osm_ucast_mgr.c
>>> @@ -998,8 +998,23 @@ int osm_ucast_mgr_process(IN osm_ucast_mgr_t *
>> p_mgr)
>>>
>>> p_osm->routing_engine_used = OSM_ROUTING_ENGINE_TYPE_NONE;
>>> while (p_routing_eng) {
>>> + if (!p_routing_eng->initialized &&
>>> + p_routing_eng->setup(p_routing_eng, p_osm)) {
>>> + OSM_LOG(p_mgr->p_log, OSM_LOG_ERROR,
>>> + "ERR 3A0F: setup of routing engine
>> \'%s\' failed\n",
>>> + p_routing_eng->name);
>>> + p_routing_eng =
>> p_routing_eng->next;
>>> + continue;
>>> + }
>>> + OSM_LOG(p_mgr->p_log, OSM_LOG_INFO,
>>> + "\'%s\' routing engine set up\n",
>> p_routing_eng->name);
>>> + p_routing_eng->initialized = 1;
>>> if (!ucast_mgr_route(p_routing_eng, p_osm))
>>> break;
>>> + /* delete unused routing engine */
>>> + if (p_routing_eng->delete)
>>> + p_routing_eng->delete(p_routing_eng->context);
>>> + p_routing_eng->initialized = 0;
>>> p_routing_eng = p_routing_eng->next;
>>> }
>>>
>>> @@ -1011,6 +1026,19 @@ int osm_ucast_mgr_process(IN osm_ucast_mgr_t *
>> p_mgr)
>>> p_osm->routing_engine_used =
>> OSM_ROUTING_ENGINE_TYPE_MINHOP;
>>> }
>>>
>>> + /* if for some reason different routing engine is used */
>>> + /* cleanup last unused routing engine */
>>> + p_routing_eng = p_osm->last_routing_engine;
>>> + if (p_routing_eng) {
>>> + if (p_routing_eng->initialized &&
>>> + p_routing_eng->delete &&
>>> + p_osm->routing_engine_used !=
>>> +
>> osm_routing_engine_type(p_routing_eng->name))
>>> +
>> p_routing_eng->delete(p_routing_eng->context);
>>> + free(p_routing_eng);
>>> + p_osm->last_routing_engine = NULL;
>>> + }
>>> +
>>> OSM_LOG(p_mgr->p_log, OSM_LOG_INFO,
>>> "%s tables configured on all switches\n",
>>>
>> osm_routing_engine_type_str(p_osm->routing_engine_used));
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-rdma"
> in
>> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2010-02-14 10:18 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-11-01 8:26 [PATCH v3] opensm: support routing engine update Slava Strebkov
[not found] ` <4AED461B.5040302-hKgKHo2Ms0F+cjeuK/JdrQ@public.gmane.org>
2009-12-06 16:03 ` Yevgeny Kliteynik
[not found] ` <4B1BD5B6.6000005-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2009-12-07 8:01 ` Slava Strebkov
[not found] ` <39C75744D164D948A170E9792AF8E7CA01F6FA2F-QfUkFaTmzUSUvQqKE/ONIwC/G2K4zDHf@public.gmane.org>
2009-12-07 8:22 ` Yevgeny Kliteynik
[not found] ` <4B1CBB34.6070708-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2009-12-07 10:27 ` Slava Strebkov
[not found] ` <39C75744D164D948A170E9792AF8E7CA01F6FA30-QfUkFaTmzUSUvQqKE/ONIwC/G2K4zDHf@public.gmane.org>
2010-02-14 10:18 ` Eli Dorfman (Voltaire) [this message]
[not found] ` <4B77CDD8.2040900-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2010-02-14 12:38 ` Yevgeny Kliteynik
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4B77CDD8.2040900@gmail.com \
--to=dorfman.eli-re5jqeeqqe8avxtiumwx3w@public.gmane.org \
--cc=kliteyn-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=slavas-smomgflXvOZWk0Htik3J/w@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox