USB stack getting confused

[O'Connor, Daniel]

> On 11 Mar 2019, at 21:27, Konstantin Belousov <kostikbel at gmail.com> wrote:
> 
> On Mon, Mar 11, 2019 at 09:27:56AM +0100, Hans Petter Selasky wrote:
>> On 3/11/19 7:27 AM, O'Connor, Daniel wrote:
>>> I just realised I can check procstat for open file, derp.
>>> 
>>> [maarsytest 6:26] ~> procstat -f 2324|grep 0.5.0| wc
>>>       64     640    4928
>>> 
>>> So I guess that is why it is giving ENOMEM, I'm leaking FDs!
> If I am interpreting the output right, it is only 634 (or close) file
> descriptors opened.

It's 64 descriptors for that particular USB device (/dev/usb/0.5.0) - I am guessing there is some hard limit in the USB stack related to that.

The system definitely isn't out of FDs generally ->
[maarsytest 12:33] ~> sysctl kern.openfiles kern.maxfiles
kern.openfiles: 386
kern.maxfiles: 521571


> Too many fds errors are ENFILE when too many file descriptors in the system
> already exist, limited by kern.maxfiles, and EMFILE when per-process
> lmit is exceeded (resource RLIMIT_NOFILE).
> 
> So your ENOMEM must come from something else.

Yes, I think it must be a limit in the USB stack.

--
Daniel O'Connor
"The nice thing about standards is that there
are so many of them to choose from."
 -- Andrew Tanenbaum