1017589 - fd is leaking when doing stability test with gecko v1.4

Status: RESOLVED INVALID

(Firefox OS Graveyard :: General, defect)

Description

[Henry Chang [:hchang]]

https://bugzilla.mozilla.org/show_bug.cgi?id=1001897#c56 and the attached log [1] at time stamp 2014-05-25 03:38:38 has shown fd was leaking during the stability test. It will cause wifi crash one day when the allocated fd is greater than 1024. 

The STR is unknown since it's a proprietary stability test. 

Since there are already a couple of graphic fd leak issues, I am sure if this is one of them or another instance. 

[1] lsof_xxxxx.txt @   https://drive.google.com/file/d/0B1cSMS8_GuAEYW1MMXZMZnRuWXc/edit?usp=sharing

Comment 1

[Henry Chang [:hchang]]

Hi Sotaro, I found you solved some graphic fd leak issues like bug 1004191. Could you please help check if this one is the same issue or any possibility else? Thanks!

Hi :tkundu, Sotaro might need your help to identify this fd leaking issue. Thanks, too!

Comment 2

[Sotaro Ikeda [:sotaro]]

I already commented in Bug 1001897 Comment 60, Bug 1001897 Comment 61. We can not say 1024 could be possible on b2g. We can not say this is a evidence of leaking. And 1024 is too low as a limiation. b2g process's file descriptor's limitation is 8192.

https://github.com/mozilla-b2g/gonk-misc/blob/master/b2g.sh#L6

Comment 3

[Sotaro Ikeda [:sotaro]]

(In reply to Sotaro Ikeda [:sotaro] from comment #2)
> We can not say 1024 could be possible on b2g.

Correction: On b2g, 1024 file descriptor could be possible in normal use cases.

Comment 4

[Sotaro Ikeda [:sotaro]]

(In reply to Henry Chang [:henry] from comment #1)
> Hi Sotaro, I found you solved some graphic fd leak issues like bug 1004191.
> Could you please help check if this one is the same issue or any possibility
> else? Thanks!

Henry, do you know if the problem happen on master? If the problem does not happen on master. Some leak fix might fix this problem.

Comment 5

[Sotaro Ikeda [:sotaro]]

Anyway 1024 limit is too low for b2g.

Comment 6

[Sotaro Ikeda [:sotaro]]

bug 1004191 could cause a file descriptor leak with bugs like Bug 988135, Bug 1006957. It seems better to be uplifted to b2g v1.4.

Comment 7

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

(In reply to Sotaro Ikeda [:sotaro] from comment #5)
> Anyway 1024 limit is too low for b2g.

I am seeing limit as 8192 for b2g process in v1.4 : https://www.codeaurora.org/cgit/quic/lf/b2g/mozilla-b2g/gonk-misc/tree/b2g.sh?h=mozilla/v1.4#n6 

I am curious to know why do you think that we have 1024 as fd limit for b2g process ?

Comment 8

[Sotaro Ikeda [:sotaro]]

(In reply to Tapas Kumar Kundu from comment #7)
> (In reply to Sotaro Ikeda [:sotaro] from comment #5)
> > Anyway 1024 limit is too low for b2g.
> 
> I am seeing limit as 8192 for b2g process in v1.4 :
> https://www.codeaurora.org/cgit/quic/lf/b2g/mozilla-b2g/gonk-misc/tree/b2g.
> sh?h=mozilla/v1.4#n6 
> 
> I am curious to know why do you think that we have 1024 as fd limit for b2g
> process ?

I just thought it seems written in Comment 0.

Comment 9

[Sotaro Ikeda [:sotaro]]

I might misunderstand the comment 0.

Comment 10

[Sotaro Ikeda [:sotaro]]

(In reply to Tapas Kumar Kundu from comment #7)
> (In reply to Sotaro Ikeda [:sotaro] from comment #5)
> > Anyway 1024 limit is too low for b2g.
> 
> I am seeing limit as 8192 for b2g process in v1.4 :
> https://www.codeaurora.org/cgit/quic/lf/b2g/mozilla-b2g/gonk-misc/tree/b2g.
> sh?h=mozilla/v1.4#n6 
> 
> I am curious to know why do you think that we have 1024 as fd limit for b2g
> process ?

Sorry, misunderstand the question. Linux's default fd limit is 1024. It seems to come from it.

Comment 11

[Sotaro Ikeda [:sotaro]]

Therefore in android, 1024 is the process's fd limit. b2g need to use more fd than android from architecture's characteristic.

Comment 12

[Sotaro Ikeda [:sotaro]]

Preparation of uplifting Bug 1004191 to b2g v1.4 is already complete. Now waiting the uplift.

Comment 13

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

(In reply to Henry Chang [:henry] from comment #1)
> Hi Sotaro, I found you solved some graphic fd leak issues like bug 1004191.
> Could you please help check if this one is the same issue or any possibility
> else? Thanks!
> 
> Hi :tkundu, Sotaro might need your help to identify this fd leaking issue.
> Thanks, too!

:henry  Sure. I am always ready to help . I have one question. Sotaro is saying that b2g process can open as many as 8192 fds . So why we are we taking it as fd leak ? I am just trying to understand fd leak issue.

Thanks for your help.

Comment 14

[Henry Chang [:hchang]]

(In reply to Tapas Kumar Kundu from comment #13)
> (In reply to Henry Chang [:henry] from comment #1)
> > Hi Sotaro, I found you solved some graphic fd leak issues like bug 1004191.
> > Could you please help check if this one is the same issue or any possibility
> > else? Thanks!
> > 
> > Hi :tkundu, Sotaro might need your help to identify this fd leaking issue.
> > Thanks, too!
> 
> :henry  Sure. I am always ready to help . I have one question. Sotaro is
> saying that b2g process can open as many as 8192 fds . So why we are we
> taking it as fd leak ? I am just trying to understand fd leak issue.
> 
> Thanks for your help.

Hi two,

Thanks for your help. 1024 fds is the limitation of wpa_supplicant because of the use of fd_set for select [1]. If there is no graphic fd leaking fd greater than 1024 is considered normal,
we might need to fix wpa_supplicant, which is not the realm of gecko.

[1] http://androidxref.com/4.4.2_r2/xref/external/wpa_supplicant_8/src/common/wpa_ctrl.c#454

Comment 15

[Michael Vines [:m1] [:evilmachines]]

(In reply to Henry Chang [:henry] from comment #14)
> we might need to fix wpa_supplicant, which is not the realm of gecko.

Thanks, we'll fix the supplicant.   With bug 1004191 coming in to v1.4 soon I'm not sure there's much else to do in this bug?

Comment 16

[Henry Chang [:hchang]]

Since 1024 would be normal fd number, there's indeed no enough evidence to show there's a fd leak issue. Close this bug and go back to the original wifi bug. Thanks!