Repository navigation
segfault in nodejs process when relaunching chromium #2872
Description
Steps to reproduce
Tell us about your environment:
- Puppeteer version: 1.5.0
- Platform / OS version: Ubuntu 18 x64
- URLs (if applicable):
- Node.js version: 10.6.0
What steps will reproduce the problem?
for(;;) {
const browser = await puppeteer.launch();
await browser.close();
}What is the expected result?
Should be able to successfully shutdown and relaunch in the same nodejs process again and again.
What happens instead?
I am experiencing this segfault inside nodejs process only some times. This repros both inside docker and on my system outside of docker (same OS / arch / node). Here is the partial stacktrace:
SegvAnalysis:
Segfault happened at: 0xe50db3 <_ZN2v88internal17ConcurrentMarking3RunEiPNS1_9TaskStateE+579>: movzbl 0xa(%r10),%eax
PC (0x00e50db3) ok
source "0xa(%r10)" (0x3f67c380a) not located in a known VMA region (needed readable region)!
destination "%eax" ok
Stack memory exhausted (SP below stack segment)
SegvReason: reading unknown VMA
Stacktrace:
#0 0x0000000000e50db3 in v8::internal::ConcurrentMarking::Run(int, v8::internal::ConcurrentMarking::TaskState*) ()
#1 0x0000000000b7eac6 in v8::internal::CancelableTask::Run() ()
#2 0x0000000000904d11 in node::BackgroundRunner(void*) ()
#3 0x00007fe931aed6db in start_thread (arg=0x7fe9306f2700) at pthread_create.c:463
pd = 0x7fe9306f2700
now = <optimized out>
unwind_buf = {cancel_jmp_buf = {{jmp_buf = {140639516698368, -8583754684441953919, 140639516696512, 0, 61660120, 140722291265856, 8588927756719225217, 8588931607480105345}, mask_was_saved = 0}}, priv = {pad = {0x0, 0x0, 0x0, 0x0}, data = {prev = 0x0, cleanup = 0x0, canceltype = 0}}}
not_first_call = <optimized out>
#4 0x00007fe93181688f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
StacktraceAddressSignature: /home/sepehr/.nvm/versions/node/v10.6.0/bin/node:11:/home/sepehr/.nvm/versions/node/v10.6.0/bin/node+a50db3:/home/sepehr/.nvm/versions/node/v10.6.0/bin/node+77eac6:/home/sepehr/.nvm/versions/node/v10.6.0/bin/node+504d11:/lib/x86_64-linux-gnu/libpthread-2.27.so+76db:/lib/x86_64-linux-gnu/libc-2.27.so+12188f
StacktraceTop:
v8::internal::ConcurrentMarking::Run(int, v8::internal::ConcurrentMarking::TaskState*) ()
v8::internal::CancelableTask::Run() ()
node::BackgroundRunner(void*) ()
start_thread (arg=0x7fe9306f2700) at pthread_create.c:463
clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
Tags: bionic
I did not post this to Chromium as the stacktrace indicates this is a crash in the nodejs process.
Activity
aslushnikov commentedon Jul 12, 2018
@sepehr-laal Interesting. The following script work fine for me on
puppeteer@next, Linux and node 10.6.0:It easily gets to 350 for me. Can you please check if puppeteer's tip-of-tree make things any better?
3p3r commentedon Jul 12, 2018
I will give @next a shot, will report back. note that this happens sporadically. It seems related to GC and having something to do with async-hooks (found this info while surfing web for similar stacktraces).
3p3r commentedon Jul 12, 2018
Update: it's hard to tell if it helped or not since it happens sporadically. I was able to reproduce it only once using the @next branch. @aslushnikov
aslushnikov commentedon Jul 12, 2018
Thanks @sepehr-laal,
This looks like a V8/nodejs bug. Would you mind reporting it over to node.js?
3p3r commentedon Jul 13, 2018
yeah that seems reasonable. will report over.
aslushnikov commentedon Jul 13, 2018
Thanks.