possible memory leak on function dlr_device_construct, libdleyna/renderer/device.c

[vpereira]

Hi, I was wondering if we wouldn't have a memory leak because we aren't freeing priv_t which could lead to denial of service (crash) as soon g_new0 fails (AFAIK g_new0 aborts when it fails)

code for reference:

dleyna-renderer/libdleyna/renderer/device.c

Line 806 in 1391290

priv_t = g_new0(prv_new_device_ct_t, 1);

[phako]

If you look at https://github.com/intel/dleyna-renderer/blob/139129086cd48879d5aeb992a841ea220cc65fd2/libdleyna/renderer/device.c#L823` the priv_t is passed as user_data to the task together with a free_func g_free. The task takes over ownership of priv_t and frees it when its done

[rbjorklin]

Probably not related to this particular function but on my machine with 16G RAM I've seen the dleyna-renderer process consume ~30% of system memory with only tmux, gedit and thunderbird running at the time. I'm guessing ~5GB of RAM use is not expected?

[phako]

it does some mmap for some use-cases. the numbers might be deceiving

[zlynx]

I had to kill -9 dleyna-renderer just now because it was using 8.5 GB actual RSS. Nothing about virtual mmap going on there.

[phako]

Were you using it at all or was it merely idling in the background?

[zlynx]

On September 18, 2019 10:21:50 AM MDT, Jens Georg ***@***.***> wrote: Were you using it at all or was it merely idling in the background?

It was just idling. I never use it, so I'm looking into how to remove it from my Fedora install. Stuff I don't use only gets one chance to screw up my laptop. -- Knowledge is Power -- Power Corrupts Study Hard -- Be Evil