[Bug 202962] [iwn] ampdu tx panic

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=202962

            Bug ID: 202962
           Summary: [iwn] ampdu tx panic
           Product: Base System
           Version: 11.0-CURRENT
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Only Me
          Priority: ---
         Component: wireless
          Assignee: freebsd-wireless at FreeBSD.org
          Reporter: henry.hu.sh at gmail.com

I've got panic twice, but I don't have a coredump. Both times the panics are
related to iwn AMPDU TX code. I've got screenshots for the 2nd time, and I've
attached them.
1st:
https://goo.gl/photos/DfBKcERDHTdnYpda6
2nd:
https://goo.gl/photos/FmuuH7viE2fATpsj8
https://goo.gl/photos/hFqNumEaapkpbHzr5
https://goo.gl/photos/3NHhUrbYH6ugkWM38

system:
FreeBSD pepsi 11.0-CURRENT FreeBSD 11.0-CURRENT #11 r287028M: Sat Aug 22
20:22:32 EDT 2015     root at pepsi:/usr/obj/usr/src/sys/MYKERNEL  amd64

For the first time, the crash happened in iwn_tx_data()+0x183, at line
4431    ring = &sc->txq[ac];
4432    if ((seqno % 256) != ring->cur) {

I don't have the register info that time, but I think that it's mostly the same
cause as the 2nd time. Possibly 'ac' is invalid.

The second crash happened in iwn_ampdu_tx_stop()+0x60, at line
7419    if (sc->txq[qid].queued != 0)

Disassemble says:
0000000000004090 <iwn_ampdu_tx_stop>:
...
    40e6:       4c 63 38                movslq (%rax),%r15
    40e9:       49 69 c7 98 28 00 00    imul   $0x2898,%r15,%rax
    40f0:       41 83 bc 05 8c 2a 00    cmpl   $0x0,0x2a8c(%r13,%rax,1)
    40f7:       00 00 

and 0x2898 = sizeof(iwn_tx_ring). %r15 should be qid, and %r15 = 0x42535f5c
which is clearly invalid(This happened to be all ASCII chars. Maybe reused?).

It seems like that someone overwrites tap->txa_private, and this caused both
panics. Or it's possible that it's freed and reused.

-- 
You are receiving this mail because:
You are the assignee for the bug.