/sbin/watchdog - yet another memory leak

[swetoast]

Still if its possible with the help of valgrind to find the leak and patch it, that would mean more stability so its better to help out and find the problem instead of just pointing out how long it would take to fill up the mem

 

[sfx2000]

I've always hated chasing down memleaks - it's akin to walking around in the parking lot to search for nickels...

But there are folks that do really get into it, and it's their efforts that should be appreciated

 

[L&LD]

Still if its possible with the help of valgrind to find the leak and patch it, that would mean more stability so its better to help out and find the problem instead of just pointing out how long it would take to fill up the mem

To just point out how long it would take to fill up the memory was not my point.

My point is that normal users shouldn't worry about this specific leak.

And all for the people that know how, to do try to find and fix it too.

 

[RMerlin]

valgrind crashes when trying to debug watchdog. A web search shows other people having the exact same crash when trying to debug an ARM binary linked with uclibc, and nobody seemed to care enough to investigate. So that's a no-go.

 

[kvic]

valgrind crashes when trying to debug watchdog. A web search shows other people having the exact same crash when trying to debug an ARM binary linked with uclibc, and nobody seemed to care enough to investigate. So that's a no-go.

valgrind doesn't work on softfloat ARM...

There used to be a patch for older version. People seem not interested in keeping it updated nor the valgrind team cared to incorporate the patch.

watchdog is not as bad as preinit. Worst case bounce it overnight with a cron job..lol.

Of course would be good to nail down the leak!

 

[sfx2000]

valgrind crashes when trying to debug watchdog. A web search shows other people having the exact same crash when trying to debug an ARM binary linked with uclibc, and nobody seemed to care enough to investigate. So that's a no-go.

Erm - that's a hint...

maybe uclibc is the problem?

 

[RMerlin]

Erm - that's a hint...

maybe uclibc is the problem?

It's crashing right at start time, so this is unrelated to the leak in any way, and is a commonly mentioned issue based on a quick Google search.

 

[sfx2000]

It's crashing right at start time, so this is unrelated to the leak in any way, and is a commonly mentioned issue based on a quick Google search.

exactly - see why the arm is crashing...

 

[RMerlin]

exactly - see why the arm is crashing...

That's the valgrind's author problem to investigate, not mine.

 

[sfx2000]

That's the valgrind's author problem to investigate, not mine.

He might appreciate your contributions, eh?

GPL for the win!

 

[kvic]

Asus ARM routers. 380 series FW. Memory leak in /sbin/watchdog process.

Some 'good' news. The memory leak seems coming from BWDPI libraries. If people do not have VP and WRS on (both TrendMicro features), then you won't see this leak.

Bad news is..seems still not fixed in the latest BWDPI libraries available from GPL 380.3264..

Perhaps @vanic (or @ASUS_ASUSWRT) could you help check and resolve it with TrendMicro?

 

[Shonk]

I havnt got the dpi engine in my build

admin@Aurora:/tmp/home/root# cat /proc/$(pidof watchdog)/status|grep VmRSS
VmRSS: 1272 kB
5 mins later
admin@Aurora:/tmp/home/root# cat /proc/$(pidof watchdog)/status|grep VmRSS
VmRSS: 1276 kB

 

[kvic]

I havnt got the dpi engine in my build

admin@Aurora:/tmp/home/root# cat /proc/$(pidof watchdog)/status|grep VmRSS
VmRSS: 1272 kB
5 mins later
admin@Aurora:/tmp/home/root# cat /proc/$(pidof watchdog)/status|grep VmRSS
VmRSS: 1276 kB

If it keeps increasing at regular intervals, there could be another leak..