Bug #5125
closed
segfault in pdch_ul_controller.c:pdch_ulc_get_node()
Added by keith about 3 years ago.
Updated about 3 years ago.
Description
Program received signal SIGSEGV, Segmentation fault.
pdch_ulc_get_node (ulc=<optimized out>, fn=fn@entry=95849) at /usr/src/debug/osmo-pcu/0.9.0+gitAUTOINC+16e1678bfc-r0.18/git/src/pdch_ul_controller.c:75
[Binary from nightly today Tue 20th April.]
osmo-pcu is most of the time crashing within a few seconds of start up
Attached is a capture of startup and debugging session. due to getting <opimized out> in gdb I went back up the trace to look for relevant data.
I think this might be relevant?:
(gdb) p *bts->trx[0].pdch[5]->ulc
$12 = {pdch = 0x2b2b2b2b, last_fn = 724249387, tree_root = {rb_node = 0x2b2b2b2b}, pool_ctx = 0x2b2b2b2b}
Files
- Status changed from New to Feedback
- Assignee set to keith
- Priority changed from High to Low
Actually.. forget it, I realise this might have to do with not running latest osmo-bts.
$12 = {pdch = 0x2b2b2b2b, last_fn = 724249387, tree_root = {rb_node = 0x2b2b2b2b}, pool_ctx = 0x2b2b2b2b}
I first paid my attention to 0x2b2b2b2b. Is it just a funny coincidence? Or do deal with a buffer overflow, and somehow osmo-pcu writes MAC block padding to the pointers? keith you can try to compile osmo-pcu with --enable-sanitize, so we can see if my assumption is correct.
Please make sure you first have master of all components.
That's running osmo-pcu-sysmo? that is, with direct phy, correct?
I just tested on my sysmobts running 20175-nightly I just updated, and everything is looking fine here, I have one MS pinging over EGPRS without problems/crashes.
Please also provide the TS layout config of your BTS/PCU.
- Status changed from Feedback to Rejected
As I said, I realised I was not running it against master osmo-bts and I saw commits there that are probably related.
Yes it was osmo-bts-sysmo with direct phy but not the latest osmo-bts.
I'm closing this as invalid, thanks for the attention, but sorry for wasting your time.
That said,I do often see errors with a variety of 1,000 MS that I don't see with one phone on my desk pinging.
If I see this again with full master, I'll open another ticket.
Also available in: Atom
PDF
