[libdvben50221] stack leaks resources on non-MMI session reconnect

[libdvben50221] stack leaks resources on non-MMI session reconnect

[Stephan Trebels]

[-- Attachment #1.1: Type: text/plain, Size: 868 bytes --]


The issue was, that LIBDVBEN50221 did not allow a CAM to re-establish
the session holding non-MMI resources if using the lowlevel interface.
The session_number was recorded on open, but not freed on close (which
IMO is an bug in the code, I attach the scaled down hg changeset). With
this change, the SMIT CAM with a showtime card works fine according to
tests so far.

The effect was, that the CAM tried to constantly close and re-open the
session and the LIBDVBEN50221 kept telling it, that the resource is
already allocated to a different session.  Additionally this caused the
library to use the _old_ session number in communications with the CAM,
which did not even exist anymore, so caused all writes of CA PMTs to
fail with EINTR.

Stephan

P.S. If there is a better place to report user-space library bugs for
linuxtv, please let me know.

[-- Attachment #1.2: dvb-apps-ca-session-leak.changeset --]
[-- Type: text/plain, Size: 1450 bytes --]

# HG changeset patch
# User Stephan Trebels <stephan@trebels.com>
# Date 1279192697 -3600
# Node ID 12928658e57ea046b35dfab485f23559b3036d42
# Parent  4ba933ff13fbd619be9bdbbf097b8dbfe0fbf679
Correctly free resources when a session is closed. This allows a CAM module to re-open a session.

diff -r 4ba933ff13fb -r 12928658e57e lib/libdvben50221/en50221_stdcam_llci.c
--- a/lib/libdvben50221/en50221_stdcam_llci.c	Sat Jul 03 15:25:16 2010 +0200
+++ b/lib/libdvben50221/en50221_stdcam_llci.c	Thu Jul 15 12:18:17 2010 +0100
@@ -374,14 +374,21 @@
 		} else if (resource_id == EN50221_APP_MMI_RESOURCEID) {
 			llci->stdcam.mmi_session_number = session_number;
 		}
+		break;
 
+	case S_SCALLBACK_REASON_CLOSE:
+		if (resource_id == EN50221_APP_MMI_RESOURCEID) {
+			llci->stdcam.mmi_session_number = -1;
+		} else if (resource_id == EN50221_APP_DATETIME_RESOURCEID) {
+			llci->datetime_session_number = -1;
+		} else if (resource_id == EN50221_APP_AI_RESOURCEID) {
+			llci->stdcam.ai_session_number = -1;
+		} else if (resource_id == EN50221_APP_CA_RESOURCEID) {
+			llci->stdcam.ca_session_number = -1;
+		} else if (resource_id == EN50221_APP_MMI_RESOURCEID) {
+			llci->stdcam.mmi_session_number = -1;
+		}
 		break;
-    case S_SCALLBACK_REASON_CLOSE:
-        if (resource_id == EN50221_APP_MMI_RESOURCEID) {
-            llci->stdcam.mmi_session_number = -1;
-        }
-
-        break;
 	}
 	return 0;
 }

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [libdvben50221] stack leaks resources on non-MMI session reconnect.

[DUBOST Brice]

On 15/07/2010 15:20, Stephan Trebels wrote:
>
> The issue was, that LIBDVBEN50221 did not allow a CAM to re-establish
> the session holding non-MMI resources if using the lowlevel interface.
> The session_number was recorded on open, but not freed on close (which
> IMO is an bug in the code, I attach the scaled down hg changeset). With
> this change, the SMIT CAM with a showtime card works fine according to
> tests so far.
>
> The effect was, that the CAM tried to constantly close and re-open the
> session and the LIBDVBEN50221 kept telling it, that the resource is
> already allocated to a different session.  Additionally this caused the
> library to use the _old_ session number in communications with the CAM,
> which did not even exist anymore, so caused all writes of CA PMTs to
> fail with EINTR.
>
> Stephan
>

Hello

Just to inform that this patch solves problems with CAM PowerCAM v4.3, 
so I think it can interest more people.

Before gnutv -cammenu (and other applications using libdvben50221) was 
returning ti;eout (-3) errors constantly after the display of the system 
IDs.

Now, the menu is working flawlessly

I cannot test the descrambling for the moment but it improved quite a 
lot the situation (communication with th CAM is now possible).

One note concerning the patch itself, the last "else if (resource_id == 
EN50221_APP_MMI_RESOURCEID)" is useless.

Best regards


-- 
Brice

A: Yes.
 >Q: Are you sure?
 >>A: Because it reverses the logical flow of conversation.
 >>>Q: Why is top posting annoying in email?

Re: [libdvben50221] stack leaks resources on non-MMI session reconnect.

[DUBOST Brice]

On 08/08/2010 23:53, DUBOST Brice wrote:
> On 15/07/2010 15:20, Stephan Trebels wrote:
>>
>> The issue was, that LIBDVBEN50221 did not allow a CAM to re-establish
>> the session holding non-MMI resources if using the lowlevel interface.
>> The session_number was recorded on open, but not freed on close (which
>> IMO is an bug in the code, I attach the scaled down hg changeset). With
>> this change, the SMIT CAM with a showtime card works fine according to
>> tests so far.
>>
>> The effect was, that the CAM tried to constantly close and re-open the
>> session and the LIBDVBEN50221 kept telling it, that the resource is
>> already allocated to a different session. Additionally this caused the
>> library to use the _old_ session number in communications with the CAM,
>> which did not even exist anymore, so caused all writes of CA PMTs to
>> fail with EINTR.
>>
>> Stephan
>>
>
> Hello
>
> Just to inform that this patch solves problems with CAM PowerCAM v4.3,
> so I think it can interest more people.
>
> Before gnutv -cammenu (and other applications using libdvben50221) was
> returning ti;eout (-3) errors constantly after the display of the system
> IDs.
>
> Now, the menu is working flawlessly
>
> I cannot test the descrambling for the moment but it improved quite a
> lot the situation (communication with th CAM is now possible).
>
> One note concerning the patch itself, the last "else if (resource_id ==
> EN50221_APP_MMI_RESOURCEID)" is useless.
>
> Best regards
>
>


Hello


After more testing this Patches allow several CAM models to work and 
don't seem to make any regression.

Is there anything to be improved/tested for having it included upstream ?

Thank you

Regards

-- 
Brice

A: Yes.
 >Q: Are you sure?
 >>A: Because it reverses the logical flow of conversation.
 >>>Q: Why is top posting annoying in email?

Re: [libdvben50221] stack leaks resources on non-MMI session reconnect.

[Stephan Trebels]

[-- Attachment #1: Type: text/plain, Size: 2170 bytes --]

Hi Brice,

I did not find more things to be changed, and it works fine for me now.
Given, that the responsiveness on this list is a bit underwhelming, I
wonder whether we can find someone with commit privileges, to push this
change.

Stephan

On Tue, 2010-11-02 at 11:15 +0100, DUBOST Brice wrote: 
> On 08/08/2010 23:53, DUBOST Brice wrote:
> > On 15/07/2010 15:20, Stephan Trebels wrote:
> >>
> >> The issue was, that LIBDVBEN50221 did not allow a CAM to re-establish
> >> the session holding non-MMI resources if using the lowlevel interface.
> >> The session_number was recorded on open, but not freed on close (which
> >> IMO is an bug in the code, I attach the scaled down hg changeset). With
> >> this change, the SMIT CAM with a showtime card works fine according to
> >> tests so far.
> >>
> >> The effect was, that the CAM tried to constantly close and re-open the
> >> session and the LIBDVBEN50221 kept telling it, that the resource is
> >> already allocated to a different session. Additionally this caused the
> >> library to use the _old_ session number in communications with the CAM,
> >> which did not even exist anymore, so caused all writes of CA PMTs to
> >> fail with EINTR.
> >>
> >> Stephan
> >>
> >
> > Hello
> >
> > Just to inform that this patch solves problems with CAM PowerCAM v4.3,
> > so I think it can interest more people.
> >
> > Before gnutv -cammenu (and other applications using libdvben50221) was
> > returning ti;eout (-3) errors constantly after the display of the system
> > IDs.
> >
> > Now, the menu is working flawlessly
> >
> > I cannot test the descrambling for the moment but it improved quite a
> > lot the situation (communication with th CAM is now possible).
> >
> > One note concerning the patch itself, the last "else if (resource_id ==
> > EN50221_APP_MMI_RESOURCEID)" is useless.
> >
> > Best regards
> >
> >
> 
> 
> Hello
> 
> 
> After more testing this Patches allow several CAM models to work and 
> don't seem to make any regression.
> 
> Is there anything to be improved/tested for having it included upstream ?
> 
> Thank you
> 
> Regards
> 


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [libdvben50221] stack leaks resources on non-MMI session reconnect.

[pierre.gronlier]

Stephan Trebels wrote, On 11/02/2010 11:38 AM:
> Hi Brice,
> 
> I did not find more things to be changed, and it works fine for me now.
> Given, that the responsiveness on this list is a bit underwhelming, I
> wonder whether we can find someone with commit privileges, to push this
> change.
> 

Hi,

I've tested this patch too and it solves problems with two different
cams. (powercam v4 and cryptoworks by philips)

Commiting it would be great.

-- 
Pierre

> Stephan
> 
> On Tue, 2010-11-02 at 11:15 +0100, DUBOST Brice wrote: 
>> On 08/08/2010 23:53, DUBOST Brice wrote:
>>> On 15/07/2010 15:20, Stephan Trebels wrote:
>>>>
>>>> The issue was, that LIBDVBEN50221 did not allow a CAM to re-establish
>>>> the session holding non-MMI resources if using the lowlevel interface.
>>>> The session_number was recorded on open, but not freed on close (which
>>>> IMO is an bug in the code, I attach the scaled down hg changeset). With
>>>> this change, the SMIT CAM with a showtime card works fine according to
>>>> tests so far.
>>>>
>>>> The effect was, that the CAM tried to constantly close and re-open the
>>>> session and the LIBDVBEN50221 kept telling it, that the resource is
>>>> already allocated to a different session. Additionally this caused the
>>>> library to use the _old_ session number in communications with the CAM,
>>>> which did not even exist anymore, so caused all writes of CA PMTs to
>>>> fail with EINTR.
>>>>
>>>> Stephan
>>>>
>>>
>>> Hello
>>>
>>> Just to inform that this patch solves problems with CAM PowerCAM v4.3,
>>> so I think it can interest more people.
>>>
>>> Before gnutv -cammenu (and other applications using libdvben50221) was
>>> returning ti;eout (-3) errors constantly after the display of the system
>>> IDs.
>>>
>>> Now, the menu is working flawlessly
>>>
>>> I cannot test the descrambling for the moment but it improved quite a
>>> lot the situation (communication with th CAM is now possible).
>>>
>>> One note concerning the patch itself, the last "else if (resource_id ==
>>> EN50221_APP_MMI_RESOURCEID)" is useless.
>>>
>>> Best regards
>>>
>>>
>>
>>
>> Hello
>>
>>
>> After more testing this Patches allow several CAM models to work and 
>> don't seem to make any regression.
>>
>> Is there anything to be improved/tested for having it included upstream ?
>>
>> Thank you
>>
>> Regards
>>
> 

Re: [libdvben50221] stack leaks resources on non-MMI session reconnect.

[Ludovic BOUE]

Hi,

I'm using this patch and it solves problems with a Powercam HD v3.1

Thank you for the commit as soon as possible. I would be easier for
libdvben50221 users.

--
Ludovic