search

node-pcap / node_pcap

libpcap bindings for node
MIT License
928 stars 253 forks source link

SEGV due to memory corruption #270

Open aanon4 opened 3 years ago

aanon4 commented 3 years ago

I really wish I could give you more, but the stack traces I get on crash mostly indicate some form of memory corruption rather than an easily reproduced issue (sorry). I have an app which creates and destroys pcap instances many times. The release of the pcap object (after it's been closed) will cause my application to SEGV at some point afterward. Sometimes, but no always, this is while the Node GC is running. The only thing I've been able to do to stop the crashes is to hang on to a reference to the binding.PcapSession internal object after I close and remove other references (this is perhaps not a surprise). Obviously having lots of those hanging about is undesirable.

Mostly filing this issue in case others are seeing this and we might compare notes.

Here's a typical crash trace:

Error in `node': double free or corruption (out): 0x00000000039bd960 ======= Backtrace: ========= /lib/x86_64-linux-gnu/libc.so.6(+0x70bfb)[0x7fd347972bfb] /lib/x86_64-linux-gnu/libc.so.6(+0x76fc6)[0x7fd347978fc6] /lib/x86_64-linux-gnu/libc.so.6(+0x7780e)[0x7fd34797980e] node(_ZN2v88internal4Heap8ScavengeEv+0x1e3a)[0xf43dba] node(_ZN2v88internal4Heap24PerformGarbageCollectionENS0_16GarbageCollectorENS_15GCCallbackFlagsE+0x3e4)[0xf49314] node(_ZN2v88internal4Heap14CollectGarbageENS0_15AllocationSpaceENS0_23GarbageCollectionReasonENS_15GCCallbackFlagsE+0x3ab)[0xf4a18b] node(_ZN2v88internal4Heap26AllocateRawWithRetryOrFailEiNS0_15AllocationSpaceENS0_19AllocationAlignmentE+0x41)[0xf4cec1] node(_ZN2v88internal7Factory30CopyJSObjectWithAllocationSiteENS0_6HandleINS0_8JSObjectEEENS2_INS0_14AllocationSiteEEE+0x95)[0xf29875] node[0x11d89c4] node[0x11de450] node(_ZN2v88internal27Runtime_CreateObjectLiteralEiPPNS0_6ObjectEPNS0_7IsolateE+0xbe)[0x11e045e] [0x3417226cfc5d]

aanon4 commented 3 years ago

Update: While not a fix as such, I'm now using a fork of this code which reuses the native pcap objects and this seems to work great. Worst case is I have N of these hanging around (where N is the number of captures I'm running in parallel).