Hi,
We’re seeing io-net with our driver (intermediate packet interceptor) crash
under a certain traffic pattern.
It has been suggested that the io-net alloc and free functions are not
re-entrant, and our use of them needs to be protected by mutex.
Anyone?
Rony Shapiro <rshapiro@everbee.com> wrote:
Hi,
We’re seeing io-net with our driver (intermediate packet interceptor) crash
under a certain traffic pattern.
It has been suggested that the io-net alloc and free functions are not
re-entrant, and our use of them needs to be protected by mutex.
No, if you stick with the “alloc()”, “alloc_up_pkt()”, “alloc_down_pkt()”,
they are all protected (by an internal mutex).
The only bad thing could happen is you double freed or sth like that.
-xtang
Hi,
Thanks for the quick reply.
It’s beginning to look like “something like that”…
io-net appears to be crashing in a call to m_free (I say “appears” because
the core file is somehow corrupt - can’t get more than a couple of frames
back).
Anyway, this seems to be after the packet leaves our driver.
Any ideas/hints/clues as to how to track this down greatly appreciated.
Cheers,
Rony
“Xiaodan Tang” <xtang@qnx.com> wrote in message
news:a31be5$l3$1@nntp.qnx.com…
Rony Shapiro <> rshapiro@everbee.com> > wrote:
Hi,
We’re seeing io-net with our driver (intermediate packet interceptor)
crash
under a certain traffic pattern.
It has been suggested that the io-net alloc and free functions are not
re-entrant, and our use of them needs to be protected by mutex.
No, if you stick with the “alloc()”, “alloc_up_pkt()”, “alloc_down_pkt()”,
they are all protected (by an internal mutex).
The only bad thing could happen is you double freed or sth like that.
-xtang