msg/async: fix bug of data type conversion when uint64_t -> int -> uint64_t

[shangfufei]

There is "uint64_t a -> int b ->uint64_t c" data type conversion in the code, when the length of "a" is greater than 32 bits, the conversion of "uint64_t a" to "int b" results in data overflow, when the length of "a" is less than 32 bits and a >= 0x80000000, change "uint64_t a" to "int b" will lead "b" to be a negative number of 32 bits, then change " int b" to " uint64_t c" , the variable "c" is not equal to the variable “a” at this situation.

Signed-off-by: shangfufei shangfufei@inspur.com

[yuyuyu101]

who is "int"

[shangfufei]

class C_tick_wakeup : public EventCallback {
AsyncConnectionRef conn;

public:
explicit C_tick_wakeup(AsyncConnectionRef c): conn(c) {}
void do_request(int fd_or_id) {
conn->tick(fd_or_id);
}
};

fd_or_id is int
time_event_next_id is uint64_t

[shangfufei]

It got assert(last_tick_id == id) in our code when EventCenter::process_time_events cb->do_request(int fd_or_id) calls AsyncConnection::tick(uint64_t id), the ceph/master removes the assert，but this issue exits.

[shangfufei]

the core infor is this:
(gdb) bt
#0 0x00007feb76d3dfcb in raise () from /lib64/libpthread.so.0
#1 0x00007feb7888513b in reraise_fatal (signum=6) at global/signal_handler.cc:135
#2 handle_fatal_signal (signum=6) at global/signal_handler.cc:209
#3
#4 0x00007feb747715f7 in raise () from /lib64/libc.so.6
#5 0x00007feb74772ce8 in abort () from /lib64/libc.so.6
#6 0x00007feb78997f6b in icfs::__icfs_assert_fail (assertion=assertion@entry=0x7feb78c3be37 "last_tick_id == id",
file=file@entry=0x7feb78c3b888 "msg/async/AsyncConnection.cc", line=line@entry=2642,
func=func@entry=0x7feb78c3c4e0 <AsyncConnection::tick(unsigned long)::PRETTY_FUNCTION> "void AsyncConnection::tick(uint64_t)")
at common/assert.cc:80
#7 0x00007feb78ae5b7e in AsyncConnection::tick (this=0x7fec37d08000, id=18446744071562068035) at msg/async/AsyncConnection.cc:2642
#8 0x00007feb78a79581 in EventCenter::process_time_events (this=this@entry=0x7feb828bd890) at msg/async/Event.cc:350
#9 0x00007feb78a7ac10 in EventCenter::process_events (this=this@entry=0x7feb828bd890, timeout_microseconds=,
timeout_microseconds@entry=30000000) at msg/async/Event.cc:423
#10 0x00007feb78ac93ba in NetworkStack::__lambda0::operator() (__closure=0x7feb828800f0) at msg/async/Stack.cc:76
#11 0x00007feb750ca220 in ?? () from /lib64/libstdc++.so.6
#12 0x00007feb76d36dc5 in start_thread () from /lib64/libpthread.so.0
#13 0x00007feb7483221d in clone () from /lib64/libc.so.6
(gdb) f 8
#8 0x00007feb78a79581 in EventCenter::process_time_events (this=this@entry=0x7feb828bd890) at msg/async/Event.cc:350
350 cb->do_request(id);
(gdb) p id
$1 = 2147483715
(gdb) f 7
#7 0x00007feb78ae5b7e in AsyncConnection::tick (this=0x7fec37d08000, id=18446744071562068035) at msg/async/AsyncConnection.cc:2642
2642 assert(last_tick_id == id);
(gdb) p id
$2 = 18446744071562068035
(gdb) p last_tick_id
$3 = 2147483715
(gdb)

[yuyuyu101]

oh, right. but we should modify the EventCallback interface instead of this fix

[shangfufei]

i think it will modify many files.

[yuyuyu101]

not so many. it's desired

[shangfufei]

ok, i will fix this bug as your method

[shangfufei]

please review @yuyuyu101

[tchaikov]

could you prefix the title of your commit message with the subcomponent your are changing ? see https://github.com/ceph/ceph/blob/master/SubmittingPatches.rst#3-describe-your-changes

[shangfufei]

ok , changed @tchaikov

[amitkumar50]

LGTM

[yuriw]

wip-yuri3-testing-2017-10-11-1645