b2g leaks kgsl memory in v2.1

RESOLVED INVALID

Status

P1
blocker
RESOLVED INVALID
4 years ago
4 years ago

People

(Reporter: tkundu, Assigned: sotaro)

Tracking

unspecified
ARM
Gonk (Firefox OS)

Firefox Tracking Flags

(blocking-b2g:2.1+)

Details

(Whiteboard: [caf priority: p2][CR 750452][MemShrink])

we are seeing following memleak in b2g process: 

==- 00051712, 1414999732, 2014-11-03 12:58:52, 2014-11-03 07:28:52 UTC, up 232334 -==
                           |       megabytes       |
           NAME   PID PPID   CPU(s) NICE   USS   PSS   RSS SWAP VSIZE OOM_ADJ USER     
            b2g  2751    1 146854.6    0 248.2 259.2 279.2  0.0 488.2       0 root     
         (Nuwa)  2757 2751    924.5    0   1.9   4.8  13.6  0.0  53.8     -16 root     
Built-in Keyboa  2953 2757    257.9   18   6.7  10.4  21.3  0.0  64.1      10 u0_a2953 
     Homescreen  3125 2751   4829.2   18  21.0  27.1  40.5  0.0  99.1       8 u0_a3125 
(Preallocated a 32701 2757      0.2   18   5.3   8.6  18.9  0.0  60.7       1 u0_a32701

System memory info:

            Total 847.5 MB
        SwapTotal   0.0 MB
     Used - cache 489.1 MB
  B2G procs (PSS) 310.2 MB
    Non-B2G procs 178.9 MB
     Free + cache 358.4 MB
             Free 132.2 MB
            Cache 226.1 MB
         SwapFree   0.0 MB


and /sys/class/kgsl/kgsl/page_alloc is 69472256 bytes = 69MB which clearly says kgsl memleak by b2g process. unfortunately we don't have DMD report for now..

We will update any DMD report shortly..

please also look into bug 1093278 which may be related to this issue.
[Blocking Requested - why for this release]:
blocking-b2g: --- → 2.1?
Flags: needinfo?(khuey)
This is going to be gfx or gaia.
Flags: needinfo?(khuey) → needinfo?(bugmail.mozilla)
Redirecting to milan as I'm not really sure who should be looking into it. I would have to learn a bunch of code before I'm able to look into this.
Flags: needinfo?(bugmail.mozilla) → needinfo?(milan)
Sotaro, can you take a look?
Flags: needinfo?(milan) → needinfo?(sotaro.ikeda.g)

Updated

4 years ago
blocking-b2g: 2.1? → 2.1+
Passing this to :sotaro for now, so we have all the blockers assigned. we'll re-assign if needed based on the investigation
Assignee: nobody → sotaro.ikeda.g
Whiteboard: [MemShrink]
(Assignee)

Comment 6

4 years ago
(In reply to Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me) from comment #0)
> and /sys/class/kgsl/kgsl/page_alloc is 69472256 bytes = 69MB which clearly
> says kgsl memleak by b2g process. unfortunately we don't have DMD report for
> now..
> 
> We will update any DMD report shortly..

Tapas, can we get the memory report?
Flags: needinfo?(sotaro.ikeda.g) → needinfo?(tkundu)
(Assignee)

Comment 7

4 years ago
(In reply to Sotaro Ikeda [:sotaro] from comment #6)
> (In reply to Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me) from
> comment #0)
> > We will update any DMD report shortly..
> 
> Tapas, can we get the memory report?

DMD is not necessary for now.
(Assignee)

Comment 8

4 years ago
Without the memory report for "about:memory", it is difficult to understand the system memory usage overview.
(Assignee)

Comment 9

4 years ago
(In reply to Sotaro Ikeda [:sotaro] from comment #6)
> (In reply to Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me) from
> comment #0)
> > and /sys/class/kgsl/kgsl/page_alloc is 69472256 bytes = 69MB which clearly
> > says kgsl memleak by b2g process. unfortunately we don't have DMD report for
> > now..
> > 
> > We will update any DMD report shortly..
> 
> Tapas, can we get the memory report?

Tapas, can you also provide the memory report when the problem does not happen?
(Assignee)

Comment 10

4 years ago
I am suspecting that canvas 2d(backed with SkiaGL) is consuming kgsl memory like Bug 1070195. In system app, "lock screen" and "Developer Hud" using canvas 2d and they do not have "willReadFrequently: true" attribute. Therefore, The canvas 2d uses SkiaGL.
(In reply to Sotaro Ikeda [:sotaro] from comment #8)
> Without the memory report for "about:memory", it is difficult to understand
> the system memory usage overview.

I also agree on this. I asked our test team to reproduce again and collect memory report. Sorry for delay. I will update here asap
Flags: needinfo?(tkundu)
Flags: needinfo?(tkundu)

Updated

4 years ago
Whiteboard: [MemShrink] → [CR 750452][MemShrink]

Updated

4 years ago
Whiteboard: [CR 750452][MemShrink] → [caf priority: p2][CR 750452][MemShrink]
We are not seeing this issue for now. We will re-open this bug with memory-report log if this issue comes again.
Flags: needinfo?(tkundu)
Status: NEW → RESOLVED
Last Resolved: 4 years ago
Resolution: --- → INVALID
You need to log in before you can comment on or make changes to this bug.