Hello,
I found that when rate_iops used with write_lat_log， it will be
high probability
that the fio will assert. The below is the core dump.
(gdb) bt
#0 0x00007f3c3ccd25f7 in raise () from /lib64/libc.so.6
#1 0x00007f3c3ccd3ce8 in abort () from /lib64/libc.so.6
#2 0x00007f3c3cccb566 in __assert_fail_base () from /lib64/libc.so.6
#3 0x00007f3c3cccb612 in __assert_fail () from /lib64/libc.so.6
#4 0x000000000042249c in get_cur_log (iolog=0x7f3c390bb8b0) at stat.c:2004
#5 __add_log_sample (iolog=iolog@entry=0x7f3c390bb8b0,
val=val@entry=1, ddir=ddir@entry=DDIR_WRITE, bs=bs@entry=4096, t=1024,
offset=offset@entry=972521472)
at stat.c:2019
#6 0x0000000000422767 in __add_log_sample (offset=972521472,
t=<optimized out>, bs=4096, ddir=DDIR_WRITE, val=1,
iolog=0x7f3c390bb8b0) at stat.c:2099
#7 add_log_sample (td=td@entry=0x7f3c1de55000, iolog=0x7f3c390bb8b0,
val=1, ddir=<optimized out>, bs=4096, offset=972521472) at stat.c:2131
#8 0x000000000042ef14 in add_log_sample (offset=<optimized out>,
bs=<optimized out>, ddir=<optimized out>, val=<optimized out>,
iolog=<optimized out>,
td=0x7f3c1de55000) at stat.c:2269
#9 add_slat_sample (td=0x7f3c1de55000, ddir=<optimized out>,
usec=<optimized out>, bs=<optimized out>, offset=<optimized out>) at
stat.c:2270
#10 0x000000000043e976 in io_u_queued (td=<optimized out>,
io_u=<optimized out>) at io_u.c:2002
#11 0x000000000045a6ba in io_queue_event (td=td@entry=0x7f3c1de55000,
io_u=io_u@entry=0x7f3c04046200, ret=ret@entry=0x7f3c0e950c30,
ddir=ddir@entry=DDIR_WRITE,
bytes_issued=bytes_issued@entry=0x7f3c0e950c60,
from_verify=from_verify@entry=0,
comp_time=comp_time@entry=0x7f3c0e950ce0)
at backend.c:545
#12 0x000000000045d4af in do_io (bytes_done=<synthetic pointer>,
td=0x7f3c1de55000) at backend.c:1010
#13 thread_main (data=<optimized out>) at backend.c:1714
#14 0x00007f3c3d269dc5 in start_thread () from /lib64/libpthread.so.0
#15 0x00007f3c3cd9321d in clone () from /lib64/libc.so.6
it points to https://github.com/axboe/fio/blob/master/stat.c#L2004
This line will judge whether nr_samples is bigger than max_samples.
I do not have time to dig into deeper. Can someone explain more about
the bug, and is there a fix to this?
Thanks a lot.
--
谦谦君子
--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
