From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; de-AT; rv:1.4) Gecko/20030922 Description of problem: If you use adsl the pppoe and pppd daemon must work together. Sometime, if /sbin/adsl-connect is reconnect the connection (break after 24 hours) the pppd is crash with the following messages: Feb 19 22:41:11 adsl01 pppoe[10239]: Sent PADT Feb 19 22:41:11 adsl01 pppd[10238]: ioctl(PPPIOCSASYNCMAP): Inappropriate ioctl for device(25) Feb 19 22:41:11 adsl01 pppd[10238]: tcflush failed: Input/output error Feb 19 22:41:11 adsl01 pppd[10238]: Couldn't release PPP unit: Invalid argument Feb 19 22:41:11 adsl01 pppd[10238]: Exit. Next crash: Feb 25 16:56:42 adsl01 pppoe[4808]: Sent PADT Feb 25 16:56:42 adsl01 pppd[4807]: ioctl(PPPIOCSASYNCMAP): Inappropriate ioctl for device(25) Feb 25 16:56:42 adsl01 pppd[4807]: tcflush failed: Input/output error Feb 25 16:56:42 adsl01 pppd[4807]: Exit. This is bad because the ADSL does not works then. This is not the only problem of pppd at RHEL3. For more informations over the ADSL problems look at bug #114875 and my next bugentry. The crash at reconnect is not every day. Sometime it works ten days without the crash and then the crash was after 48 hours. Version-Release number of selected component (if applicable): ppp-2.4.1-14 How reproducible: Sometimes Steps to Reproduce: 1.use adsl with the reconnect after break option 2.wait for reconnect after 24 hours 3. Actual Results: pppd does not work stabil Expected Results: should work stabil Additional info:
Created attachment 98100 [details] reconnect messages during adsl line was down at provider messages from pppd reconnect during adsl line was down at provider
Created attachment 98156 [details] Sysreport adsl01
There are different versions for ppp in RHEL3. # strings lib/modules/2.4.21-9.0.1.EL/kernel/drivers/net/ppp_generic.o | grep version Kernelmodules are "PPP generic driver version 2.4.2" The RPM ppp which contains the pppd is version 2.4.1. # rpm -qa ppp ppp-2.4.1-14 You see it also in the SRPM. I build a new ppp RPM based on ppp version 2.4.2 self. I do not see error messages in /var/log/messages. pppd works correct since 14 days.
Created attachment 99072 [details] A patch which might fix the problem ppp-2.4.2 allows error returns of ENOTTY in some places where ppp-2.4.1 dies with a call to fatal(). This patch changes ppp-2.4.1 to allow ENOTTY there too. It also fixes a crash in option parsing. I cannot test this patch unfortunately due to lack of the necessary hardware.
Please have a look at http://people.redhat.com/twoerner/SRPMS/3.0E/ppp-2.4.1-14.1.src.rpm http://people.redhat.com/twoerner/RPMS/3.0E/ppp-2.4.1-14.1.i386.rpm
First results for test rp-pppoe-3.5-4.1.i386.rpm and ppp-2.4.1-14.1.i386.rpm together are: 1. Dial on Demand - works now, the option idle=yes do not longer comes from rp-pppoe - this mistake was the reason for pppd-crash without a message you can see the invalid option (I found this with my own ppp-2.4.2 rpm) 2. use adsl with the reconnect after break option: - in my first tests with own simulation the reconnet the pppd works correct - also the last two reconnets after 24 hours were without errors At this time I have only my own firewall-script and start/stop squid in ip-up.local and ip-down.local activ. Squid has no traffic and also my VPN-IPSec is not activ. Both I will include tomorrow to see in the next two weeks if the pppd-2.4.1 works really correct like ppp-2.4.2 now. Uwe
Block for Bug #114875 and #116727 not longer needed. I have no access to Bug #116727 but I think Bug #116927 is the correct number. For the reported mistakes is the pppd the reason. The fixes in rp-pppoe-3.5-4.1.i386.rpm are o.k.. Uwe
The ppp-2.4.1-14.1.i386.rpm do not fix the problem. You can get detailed information from RHEL hotline. Look for issue-tracker issue 35179. Uwe
Pushing back to assigned, as reporter still appears to be experiencing issues.
Ok, what exactly is now working and what is not?
I start a new test with ppp-2.4.1-14.1.i386.rpm yesterday. It is on a productiv machine! It is correct, the last report in issue-tracker issue 35179 was a contretemps. It is not normal, that during the if-up.local script the ADSL line breaks again. But the same if-up.local and if-down.local scripts never abort short with my own ppp-2.4.2 rpm. I have only seen the problem with ppp-2.4.1 rpm from RHEL. Correct, bevor the fix. If I have the same breaks again now, I can only change to my own ppp-2.4.2 rpm!? Uwe
Is "ioctl(PPPIOCSASYNCMAP): Inappropriate ioctl for device(25)" fixed for you now?
No problem since 2004-08-22. Only the U3 kernel update interrupt the long time test. Uwe
Closing.