9bc2cbb007
We watched a rte panic of mbuf_autotest in our qualcomm arm64 server
(Amberwing).
Root cause:
In __rte_ring_move_cons_head()
...
do {
/* Restore n as it may change every loop */
n = max;
*old_head = r->cons.head; //1st load
const uint32_t prod_tail = r->prod.tail; //2nd load
In weak memory order architectures (powerpc,arm), the 2nd load might be
reodered before the 1st load, that makes *entries is bigger than we wanted.
This nasty reording messed enque/deque up.
cpu1(producer) cpu2(consumer) cpu3(consumer)
load r->prod.tail
in enqueue:
load r->cons.tail
load r->prod.head
store r->prod.tail
load r->cons.head
load r->prod.tail
...
store r->cons.{head,tail}
load r->cons.head
Then, r->cons.head will be bigger than prod_tail, then make *entries very
big and the consumer will go forward incorrectly.
After this patch, the old cons.head will be recaculated after failure of
rte_atomic32_cmpset
There is no such issue on X86, because X86 is strong memory order model.
But rte_smp_rmb() doesn't have impact on runtime performance on X86, so
keep the same code without architectures specific concerns.
Fixes:
|
||
|---|---|---|
| app | ||
| buildtools | ||
| config | ||
| devtools | ||
| doc | ||
| drivers | ||
| examples | ||
| lib | ||
| mk | ||
| pkg | ||
| test | ||
| usertools | ||
| .gitattributes | ||
| .gitignore | ||
| GNUmakefile | ||
| LICENSE.GPL | ||
| LICENSE.LGPL | ||
| MAINTAINERS | ||
| Makefile | ||
| README | ||
DPDK is a set of libraries and drivers for fast packet processing. It supports many processor architectures and both FreeBSD and Linux. The DPDK uses the Open Source BSD license for the core libraries and drivers. The kernel components are GPLv2 licensed. Please check the doc directory for release notes, API documentation, and sample application information. For questions and usage discussions, subscribe to: users@dpdk.org Report bugs and issues to the development mailing list: dev@dpdk.org