* [PATCH] [dvbv5-scan] wait no more than timeout when scanning
@ 2016-07-22 17:54 Abylay Ospan
2016-07-22 19:18 ` Mauro Carvalho Chehab
0 siblings, 1 reply; 5+ messages in thread
From: Abylay Ospan @ 2016-07-22 17:54 UTC (permalink / raw)
To: Mauro Carvalho Chehab, linux-media; +Cc: Abylay Ospan
some frontends (mentioned on lgdt3306a) wait timeout inside code like:
for (i = 20; i > 0; i--) {
msleep(50);
If there is no-LOCK then dvbv5-scan spent a lot of time (doing 40x calls).
This patch introduce timeout which 4 sec * multiply. So we do not wait more
than 4 sec (or so) if no-LOCK.
Signed-off-by: Abylay Ospan <aospan@netup.ru>
---
utils/dvb/dvbv5-scan.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/utils/dvb/dvbv5-scan.c b/utils/dvb/dvbv5-scan.c
index 689bc0b..1fc33d7 100644
--- a/utils/dvb/dvbv5-scan.c
+++ b/utils/dvb/dvbv5-scan.c
@@ -182,12 +182,23 @@ static int print_frontend_stats(struct arguments *args,
return 0;
}
+/* return timestamp in msec */
+uint64_t get_timestamp()
+{
+ struct timeval now;
+ gettimeofday(&now, 0);
+ return now.tv_sec * 1000 + now.tv_usec/1000;
+}
+
static int check_frontend(void *__args,
struct dvb_v5_fe_parms *parms)
{
struct arguments *args = __args;
int rc, i;
fe_status_t status;
+ uint64_t start = get_timestamp();
+ /* msec timeout by default 4 sec * multiply */
+ uint64_t timeout = args->timeout_multiply * 4 * 1000;
args->n_status_lines = 0;
for (i = 0; i < args->timeout_multiply * 40; i++) {
@@ -203,6 +214,10 @@ static int check_frontend(void *__args,
print_frontend_stats(args, parms);
if (status & FE_HAS_LOCK)
break;
+
+ if ((get_timestamp() - start) > timeout)
+ break;
+
usleep(100000);
};
--
2.7.4
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] [dvbv5-scan] wait no more than timeout when scanning
2016-07-22 17:54 [PATCH] [dvbv5-scan] wait no more than timeout when scanning Abylay Ospan
@ 2016-07-22 19:18 ` Mauro Carvalho Chehab
2016-07-25 4:01 ` Abylay Ospan
2016-07-25 18:47 ` Abylay Ospan
0 siblings, 2 replies; 5+ messages in thread
From: Mauro Carvalho Chehab @ 2016-07-22 19:18 UTC (permalink / raw)
To: Abylay Ospan; +Cc: linux-media
Hi Abylay,
Em Fri, 22 Jul 2016 13:54:37 -0400
Abylay Ospan <aospan@netup.ru> escreveu:
> some frontends (mentioned on lgdt3306a) wait timeout inside code like:
> for (i = 20; i > 0; i--) {
> msleep(50);
>
> If there is no-LOCK then dvbv5-scan spent a lot of time (doing 40x calls).
> This patch introduce timeout which 4 sec * multiply. So we do not wait more
> than 4 sec (or so) if no-LOCK.
>
> Signed-off-by: Abylay Ospan <aospan@netup.ru>
> ---
> utils/dvb/dvbv5-scan.c | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
> diff --git a/utils/dvb/dvbv5-scan.c b/utils/dvb/dvbv5-scan.c
> index 689bc0b..1fc33d7 100644
> --- a/utils/dvb/dvbv5-scan.c
> +++ b/utils/dvb/dvbv5-scan.c
> @@ -182,12 +182,23 @@ static int print_frontend_stats(struct arguments *args,
> return 0;
> }
>
> +/* return timestamp in msec */
> +uint64_t get_timestamp()
> +{
> + struct timeval now;
> + gettimeofday(&now, 0);
> + return now.tv_sec * 1000 + now.tv_usec/1000;
This is not good, as gettimeofday() is not monotonic, and may be affected
by clock adjustments.
IMHO, the best would be to adjust the do_timeout() to handle
args->timeout_multiply.
Regards,
Mauro
> +}
> +
> static int check_frontend(void *__args,
> struct dvb_v5_fe_parms *parms)
> {
> struct arguments *args = __args;
> int rc, i;
> fe_status_t status;
> + uint64_t start = get_timestamp();
> + /* msec timeout by default 4 sec * multiply */
> + uint64_t timeout = args->timeout_multiply * 4 * 1000;
>
> args->n_status_lines = 0;
> for (i = 0; i < args->timeout_multiply * 40; i++) {
> @@ -203,6 +214,10 @@ static int check_frontend(void *__args,
> print_frontend_stats(args, parms);
> if (status & FE_HAS_LOCK)
> break;
> +
> + if ((get_timestamp() - start) > timeout)
> + break;
> +
> usleep(100000);
It would also make sense to remove the usleep here and
use something else that would be checking timeout_flag,
like:
for (i = 1; i < 100; i++) {
if (timeout_flag)
break;
usleep(1000);
}
--
Thanks,
Mauro
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] [dvbv5-scan] wait no more than timeout when scanning
2016-07-22 19:18 ` Mauro Carvalho Chehab
@ 2016-07-25 4:01 ` Abylay Ospan
2016-07-25 18:47 ` Abylay Ospan
1 sibling, 0 replies; 5+ messages in thread
From: Abylay Ospan @ 2016-07-25 4:01 UTC (permalink / raw)
To: Mauro Carvalho Chehab; +Cc: linux-media
Hi Mauro,
I have checked 'do_timeout' - it's bit different that we want. I just
updated patch to use 'clock_gettime(CLOCK_MONOTONIC' to avoid
timestamps 'rollup'.
Hope it's ok. The problem is really annoying because scanning takes a
lot of time.
Without this patch command 'dvbv5-scan
us-ATSC-center-frequencies-8VSB' takes about 2 hours.
with patch it takes just 6min 45sec - which better but has a room for
optimization inside lgdt3306a.c driver - it waits about 4 sec for each
read status which I think is too long. If we have signal we should
lock in about 100-500 msec. I will check this question later.
2016-07-22 15:18 GMT-04:00 Mauro Carvalho Chehab <mchehab@osg.samsung.com>:
> Hi Abylay,
>
> Em Fri, 22 Jul 2016 13:54:37 -0400
> Abylay Ospan <aospan@netup.ru> escreveu:
>
>> some frontends (mentioned on lgdt3306a) wait timeout inside code like:
>> for (i = 20; i > 0; i--) {
>> msleep(50);
>>
>> If there is no-LOCK then dvbv5-scan spent a lot of time (doing 40x calls).
>> This patch introduce timeout which 4 sec * multiply. So we do not wait more
>> than 4 sec (or so) if no-LOCK.
>>
>> Signed-off-by: Abylay Ospan <aospan@netup.ru>
>> ---
>> utils/dvb/dvbv5-scan.c | 15 +++++++++++++++
>> 1 file changed, 15 insertions(+)
>>
>> diff --git a/utils/dvb/dvbv5-scan.c b/utils/dvb/dvbv5-scan.c
>> index 689bc0b..1fc33d7 100644
>> --- a/utils/dvb/dvbv5-scan.c
>> +++ b/utils/dvb/dvbv5-scan.c
>> @@ -182,12 +182,23 @@ static int print_frontend_stats(struct arguments *args,
>> return 0;
>> }
>>
>> +/* return timestamp in msec */
>> +uint64_t get_timestamp()
>> +{
>> + struct timeval now;
>> + gettimeofday(&now, 0);
>> + return now.tv_sec * 1000 + now.tv_usec/1000;
>
> This is not good, as gettimeofday() is not monotonic, and may be affected
> by clock adjustments.
>
> IMHO, the best would be to adjust the do_timeout() to handle
> args->timeout_multiply.
>
> Regards,
> Mauro
>
>> +}
>> +
>> static int check_frontend(void *__args,
>> struct dvb_v5_fe_parms *parms)
>> {
>> struct arguments *args = __args;
>> int rc, i;
>> fe_status_t status;
>> + uint64_t start = get_timestamp();
>> + /* msec timeout by default 4 sec * multiply */
>> + uint64_t timeout = args->timeout_multiply * 4 * 1000;
>>
>> args->n_status_lines = 0;
>> for (i = 0; i < args->timeout_multiply * 40; i++) {
>> @@ -203,6 +214,10 @@ static int check_frontend(void *__args,
>> print_frontend_stats(args, parms);
>> if (status & FE_HAS_LOCK)
>> break;
>> +
>> + if ((get_timestamp() - start) > timeout)
>> + break;
>> +
>> usleep(100000);
>
> It would also make sense to remove the usleep here and
> use something else that would be checking timeout_flag,
> like:
>
> for (i = 1; i < 100; i++) {
> if (timeout_flag)
> break;
> usleep(1000);
> }
>
>
> --
> Thanks,
> Mauro
--
Abylay Ospan,
NetUP Inc.
http://www.netup.tv
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] [dvbv5-scan] wait no more than timeout when scanning
2016-07-22 19:18 ` Mauro Carvalho Chehab
2016-07-25 4:01 ` Abylay Ospan
@ 2016-07-25 18:47 ` Abylay Ospan
1 sibling, 0 replies; 5+ messages in thread
From: Abylay Ospan @ 2016-07-25 18:47 UTC (permalink / raw)
To: Mauro Carvalho Chehab; +Cc: linux-media
Hi Mauro,
I have sent two patches to ML. One is for dvbv5-scan (v3 of this
patch) and second for lgdt3306a driver. Now I have achieved 4 minutes
for full scan (file us-ATSC-center-frequencies-8VSB with 68
frequencies inside). This is much better than 2 hours before :)
But need to test this patches if possible. Who works with lgdt3306a -
please test. Hope nothing is broken.
Thanks !
2016-07-22 15:18 GMT-04:00 Mauro Carvalho Chehab <mchehab@osg.samsung.com>:
> Hi Abylay,
>
> Em Fri, 22 Jul 2016 13:54:37 -0400
> Abylay Ospan <aospan@netup.ru> escreveu:
>
>> some frontends (mentioned on lgdt3306a) wait timeout inside code like:
>> for (i = 20; i > 0; i--) {
>> msleep(50);
>>
>> If there is no-LOCK then dvbv5-scan spent a lot of time (doing 40x calls).
>> This patch introduce timeout which 4 sec * multiply. So we do not wait more
>> than 4 sec (or so) if no-LOCK.
>>
>> Signed-off-by: Abylay Ospan <aospan@netup.ru>
>> ---
>> utils/dvb/dvbv5-scan.c | 15 +++++++++++++++
>> 1 file changed, 15 insertions(+)
>>
>> diff --git a/utils/dvb/dvbv5-scan.c b/utils/dvb/dvbv5-scan.c
>> index 689bc0b..1fc33d7 100644
>> --- a/utils/dvb/dvbv5-scan.c
>> +++ b/utils/dvb/dvbv5-scan.c
>> @@ -182,12 +182,23 @@ static int print_frontend_stats(struct arguments *args,
>> return 0;
>> }
>>
>> +/* return timestamp in msec */
>> +uint64_t get_timestamp()
>> +{
>> + struct timeval now;
>> + gettimeofday(&now, 0);
>> + return now.tv_sec * 1000 + now.tv_usec/1000;
>
> This is not good, as gettimeofday() is not monotonic, and may be affected
> by clock adjustments.
>
> IMHO, the best would be to adjust the do_timeout() to handle
> args->timeout_multiply.
>
> Regards,
> Mauro
>
>> +}
>> +
>> static int check_frontend(void *__args,
>> struct dvb_v5_fe_parms *parms)
>> {
>> struct arguments *args = __args;
>> int rc, i;
>> fe_status_t status;
>> + uint64_t start = get_timestamp();
>> + /* msec timeout by default 4 sec * multiply */
>> + uint64_t timeout = args->timeout_multiply * 4 * 1000;
>>
>> args->n_status_lines = 0;
>> for (i = 0; i < args->timeout_multiply * 40; i++) {
>> @@ -203,6 +214,10 @@ static int check_frontend(void *__args,
>> print_frontend_stats(args, parms);
>> if (status & FE_HAS_LOCK)
>> break;
>> +
>> + if ((get_timestamp() - start) > timeout)
>> + break;
>> +
>> usleep(100000);
>
> It would also make sense to remove the usleep here and
> use something else that would be checking timeout_flag,
> like:
>
> for (i = 1; i < 100; i++) {
> if (timeout_flag)
> break;
> usleep(1000);
> }
>
>
> --
> Thanks,
> Mauro
--
Abylay Ospan,
NetUP Inc.
http://www.netup.tv
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] [dvbv5-scan] wait no more than timeout when scanning
@ 2016-07-25 3:35 Abylay Ospan
0 siblings, 0 replies; 5+ messages in thread
From: Abylay Ospan @ 2016-07-25 3:35 UTC (permalink / raw)
To: Mauro Carvalho Chehab, linux-media; +Cc: Abylay Ospan
some frontends (mentioned on lgdt3306a) wait timeout inside code like:
for (i = 20; i > 0; i--) {
msleep(50);
If there is no-LOCK then dvbv5-scan spent a lot of time (doing 40x calls).
This patch introduce timeout which 4 sec * multiply. So we do not wait more
than 4 sec (or so) if no-LOCK.
CLOCK_MONOTONIC is used so we don't care about timestamps "rollup"
Signed-off-by: Abylay Ospan <aospan@netup.ru>
---
utils/dvb/dvbv5-scan.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/utils/dvb/dvbv5-scan.c b/utils/dvb/dvbv5-scan.c
index 689bc0b..f6bb3fc 100644
--- a/utils/dvb/dvbv5-scan.c
+++ b/utils/dvb/dvbv5-scan.c
@@ -182,12 +182,23 @@ static int print_frontend_stats(struct arguments *args,
return 0;
}
+/* return timestamp in msec */
+uint64_t get_timestamp()
+{
+ struct timespec now;
+ clock_gettime(CLOCK_MONOTONIC, &now);
+ return now.tv_sec * 1000 + now.tv_nsec/1000000;
+}
+
static int check_frontend(void *__args,
struct dvb_v5_fe_parms *parms)
{
struct arguments *args = __args;
int rc, i;
fe_status_t status;
+ uint64_t start = get_timestamp();
+ /* msec timeout by default 4 sec * multiply */
+ uint64_t timeout = args->timeout_multiply * 4 * 1000;
args->n_status_lines = 0;
for (i = 0; i < args->timeout_multiply * 40; i++) {
@@ -203,6 +214,10 @@ static int check_frontend(void *__args,
print_frontend_stats(args, parms);
if (status & FE_HAS_LOCK)
break;
+
+ if ((get_timestamp() - start) > timeout)
+ break;
+
usleep(100000);
};
--
2.7.4
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2016-07-25 18:47 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-07-22 17:54 [PATCH] [dvbv5-scan] wait no more than timeout when scanning Abylay Ospan
2016-07-22 19:18 ` Mauro Carvalho Chehab
2016-07-25 4:01 ` Abylay Ospan
2016-07-25 18:47 ` Abylay Ospan
2016-07-25 3:35 Abylay Ospan
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.