6:42 p.m.
Hi all,
I'm running into a problem with AT command read process and I need some
help because I don't fully understand the code in this case.
Ofono freezes in one particular systematic case when I receive "\r\nOK\r\n"
from the modem port (without quotes).
As I understood the method gatchat/gatchat.c::new_bytes() is called when
something was read from gatchat/gatio.c::received_data(), and the
function p->syntax->feed returns a result based on the syntax structure.
I made some debug and I found out that, in the failing case, syntax->state
is 2, it means that it is trying to read a line, but the process fails
because it doesn't find the ' " ' character.
I don't understand where the variable sysntax->state was set to 2. It
doesn't make sense to me.
I tried to search the code for all possible points where the variable is
set but I couldn't find the correct one.
Does anyone have a suggestion about where should I look?
Thank you all.
Enrico
12:04 a.m.
Hi Enrico,
On 05/05/2016 03:42 AM, Enrico Sau wrote:
Hi all,
I'm running into a problem with AT command read process and I need some
help because I don't fully understand the code in this case.
Ofono freezes in one particular systematic case when I receive
"\r\nOK\r\n" from the modem port (without quotes).
As I understood the method gatchat/gatchat.c::new_bytes() is called when
something was read from gatchat/gatio.c::received_data(), and the
function p->syntax->feed returns a result based on the syntax structure.
I made some debug and I found out that, in the failing case,
syntax->state is 2, it means that it is trying to read a line, but the
process fails because it doesn't find the ' " ' character.
What syntax is being used?
I don't understand where the variable sysntax->state was set
to 2. It
doesn't make sense to me.
I tried to search the code for all possible points where the variable is
set but I couldn't find the correct one.
How does the rest of the log look like?
Does anyone have a suggestion about where should I look?
Thank you all.
Enrico
Regards,
-Denis
12:51 a.m.
Hi Denis,
What syntax is being used?
I'm using permissive syntax.
How does the rest of the log look like?
The rest of the log look fine to me, I had to add prints to understand
why it was blocking.
Anyway, I managed to solve the problem with a simple yet potentially
dangerous fix: I deleted the if statement in the
GSM_PERMISSIVE_STATE_RESPONSE_STRING case, so, it changed from:
gatchat/gatsyntax.c (row 309 more or less)
case GSM_PERMISSIVE_STATE_RESPONSE_STRING:
if (byte == '"')
syntax->state = GSM_PERMISSIVE_STATE_RESPONSE;
break;
To:
case GSM_PERMISSIVE_STATE_RESPONSE_STRING:
syntax->state = GSM_PERMISSIVE_STATE_RESPONSE;
break;
With this fix everything seem to work in my case.
Do you think it is dangerous in any way?
I still don't understand where the "syntax->state =
GSM_PERMISSIVE_STATE_RESPONSE_STRING"
comes from since the state seems to be managed only in this function. I'm
thinking about some memory corruption but I'm not able to find why.
Thank you,
Enrico
1:10 a.m.
Hi Enrico,
Anyway, I managed to solve the problem with a simple yet potentially
dangerous fix: I deleted the if statement in the
GSM_PERMISSIVE_STATE_RESPONSE_STRING case, so, it changed from:
gatchat/gatsyntax.c (row 309 more or less)
case GSM_PERMISSIVE_STATE_RESPONSE_STRING:
if (byte == '"')
syntax->state = GSM_PERMISSIVE_STATE_RESPONSE;
break;
To:
case GSM_PERMISSIVE_STATE_RESPONSE_STRING:
syntax->state = GSM_PERMISSIVE_STATE_RESPONSE;
break;
With this fix everything seem to work in my case.
Do you think it is dangerous in any way?
Yes. This particular logic enables strings (e.g. text enclosed in
quotes) to contain various special characters that would otherwise be
interpreted differently.
It looks like you have an unterminated " in one of the AT command
responses...
Regards,
-Denis
1:52 a.m.
Hi Denis,
This particular logic enables strings (e.g. text enclosed in quotes)
to
contain various special characters that would otherwise be interpreted
differently.
Ok, understood. I'll try to find another solution.
It looks like you have an unterminated " in one of the AT command
responses...
I will check for this, but what I think is that the state variable is set
outside the syntax state machine, because in my added prints I see that it
is already set to GSM_PERMISSIVE_STATE_RESPONSE_STRING before any parsing.
Does it look strange to you?
Thank you,
Enrico
8:06 p.m.
Hi Denis,
discard my last mail please, I went back some step and I found that in the
modem bus there are spare characters which may get the state machine
confused.
I'm thinking abut solving the problem from the plugin because this happens
when I insert the sim.
Is there a safe way to reset the modem at channel including the syntax from
the plugin?
Thank you,
Enrico
1:48 a.m.
Hi Enrico,
On 05/06/2016 05:06 AM, Enrico Sau wrote:
Hi Denis,
discard my last mail please, I went back some step and I found that in
the modem bus there are spare characters which may get the state
machine confused.
Do you have a log handy?
I'm thinking abut solving the problem from the plugin because this
happens when I insert the sim.
Is there a safe way to reset the modem at channel including the syntax
from the plugin?
Not really, but you can try writing your own syntax if the garbage the
modem sends is not random.
Regards,
-Denis
5:43 p.m.
Hi Denis,
Do you have a log handy?
Yes, I attached a log with the following operations:
- enable modem
- activate context
- extract SIM
- insert SIM
- activate context
Not really, but you can try writing your own syntax if the garbage
the
modem sends is not random.
This is a good idea, I will check the randomness of the garbage ad see if
I can do it.
Thank you.
Enrico
6:22 p.m.
Hi Denis,
I managed to handle the garbage adding a new syntax which is the copy/paste
of gsm_permissive plus a new state.
Here is how the new feed looks:
static GAtSyntaxResult gsm_telit_feed(GAtSyntax *syntax, const char
*bytes, gsize *len)
{
gsize i = 0;
GAtSyntaxResult res = G_AT_SYNTAX_RESULT_UNSURE;
while (i < *len) {
char byte = bytes[i];
switch (syntax->state) {
case GSM_TELIT_STATE_IDLE:
if (byte == '\r' || byte == '\n')
/* ignore */;
else if (byte == '>')
syntax->state = GSM_TELIT_STATE_PROMPT;
else if (byte == '~')
syntax->state = GSM_TELIT_STATE_GARBAGE;
else if (byte == '"')
syntax->state = GSM_TELIT_STATE_RESPONSE_STRING;
else
syntax->state = GSM_TELIT_STATE_RESPONSE;
break;
case GSM_TELIT_STATE_RESPONSE:
if (byte == '\r') {
syntax->state = GSM_TELIT_STATE_IDLE;
i += 1;
res = G_AT_SYNTAX_RESULT_LINE;
goto out;
} else if (byte == '"')
syntax->state =
GSM_TELIT_STATE_RESPONSE_STRING;
break;
case GSM_TELIT_STATE_RESPONSE_STRING:
if (byte == '"')
syntax->state = GSM_TELIT_STATE_RESPONSE;
break;
case GSM_TELIT_STATE_GUESS_PDU:
if (byte != '\r' && byte != '\n')
syntax->state = GSM_TELIT_STATE_PDU;
break;
case GSM_TELIT_STATE_PDU:
if (byte == '\r') {
syntax->state = GSM_TELIT_STATE_IDLE;
i += 1;
res = G_AT_SYNTAX_RESULT_PDU;
goto out;
}
break;
case GSM_TELIT_STATE_PROMPT:
if (byte == ' ') {
syntax->state = GSM_TELIT_STATE_IDLE;
i += 1;
res = G_AT_SYNTAX_RESULT_PROMPT;
goto out;
}
syntax->state = GSM_TELIT_STATE_RESPONSE;
return G_AT_SYNTAX_RESULT_UNSURE;
case GSM_TELIT_STATE_GUESS_SHORT_PROMPT:
if (byte == '\n')
/* ignore */;
else if (byte == '\r')
syntax->state = GSM_TELIT_STATE_SHORT_PROMPT;
else
syntax->state = GSM_TELIT_STATE_RESPONSE;
break;
case GSM_TELIT_STATE_SHORT_PROMPT:
if (byte == '\n') {
syntax->state = GSM_TELIT_STATE_IDLE;
i += 1;
res = G_AT_SYNTAX_RESULT_PROMPT;
goto out;
}
syntax->state = GSM_TELIT_STATE_RESPONSE;
return G_AT_SYNTAX_RESULT_UNSURE;
case GSM_TELIT_STATE_GARBAGE:
if (byte == '~')
syntax->state = GSM_TELIT_STATE_RESPONSE;
break;
default:
break;
};
i += 1;
}
out:
*len = i;
return res;
}
Enrico
12:19 a.m.
Hi Enrico,
On 05/09/2016 02:43 AM, Enrico Sau wrote:
ofonod[1306]: Aux: < #QSS: 0\r\n
ofonod[1306]: 09:33:54|plugins/he910.c[173]:he910_qss_notify() 0x88ef560
ofonod[1306]: 09:33:54|plugins/he910.c[131]:switch_sim_state_status() 0x88ef560, SIM
status: 0
ofonod[1306]: 09:33:57|src/modem.c[496]:modem_change_state() old state: 3, new state: 1
ofonod[1306]: 09:33:57|src/modem.c[409]:flush_atoms()
ofonod[1306]: 09:34:04|src/gprs.c[2279]:gprs_context_unregister() 0x88eef68, 0x88ee908
ofonod[1306]: 09:34:04|src/gprs.c[2413]:gprs_context_remove() atom: 0x88eef88
ofonod[1306]: 09:33:42|drivers/atmodem/gprs-context.c[420]:at_gprs_context_remove()
ofonod[1306]: 09:34:04|src/gprs.c[2644]:gprs_unregister() 0x88ee908
ofonod[1306]: 09:33:58|src/network.c[1173]:__ofono_netreg_remove_status_watch()
0x88eb2d0
ofonod[1306]: 09:34:04|src/gprs.c[2683]:gprs_remove() atom: 0x88ee970
ofonod[1306]: 09:34:02|src/sim.c[2714]:ofono_sim_remove_spn_watch() 0x88ed6f0
ofonod[1306]: 09:33:58|src/network.c[1836]:netreg_remove() atom: 0x88eb338
ofonod[1306]: 09:34:09|src/telit-data-network.c[323]:telit_data_network_remove() atom:
0x88eb1c8
ofonod[1306]: 09:34:09|src/telit-urc.c[668]:telit_urc_remove() atom: 0x88eb0a8
ofonod[1306]: 09:34:05|src/radio-settings.c[657]:radio_settings_remove() atom: 0x88ee578
ofonod[1306]: 09:33:28|plugins/udevng.c[1009]:remove_device()
/sys/devices/virtual/net/ppp0
ofonod[1306]: Aux: < \r\n
ofonod[1306]: Aux: < +CGEV: NW DEACT IP, "2.194.193.31", 1\r\n
ofonod[1306]: Modem: < ~\377}#\300!}%}"} }$Y(~
Now I get what is happening. You're on a PPP link and pull the SIM. So
oFono tears everything down, but this results in the GAtChat going from
PPP mode to normal mode too early.
We then receive the final HDLC frame, and that confuses the syntax.
GSMv1 syntax already has code to handle this, but looks like the
permissive one doesn't.
ofonod[1306]: Modem: < \r\nNO CARRIER\r\n
ofonod[1306]: Modem: < \r\n
ofonod[1306]: Aux: < \r\n
Regards,
-Denis
2323
days inactive
2327
days old
9 comments
2 participants
participants (2)
-
Denis Kenzior -
Enrico Sau