[dpdk-dev] [ovs-dev] OvS DPDK crash when HPET timer enabled

Mon Apr 27 13:06:38 CEST 2020

On Wed, Apr 15, 2020 at 5:27 AM Ravi Kerur <rkerur at gmail.com> wrote:
>
> Hi Tonghao, Thanks, your patch works fine with HPET initialization. I will look into DPDK discussion which got started with your patch.
[update] patch applied:
http://git.dpdk.org/next/dpdk-next-net/commit/?id=9b41c276fd35fe3c0ed3fb0d9cf98a83a58b31b2
> Thanks,
> Ravi
>
>
> On Mon, Apr 13, 2020 at 5:33 PM Tonghao Zhang <xiangxia.m.yue at gmail.com> wrote:
>>
>> On Tue, Apr 14, 2020 at 7:20 AM Ilya Maximets <i.maximets at ovn.org> wrote:
>> >
>> > On 4/13/20 5:27 PM, Ravi Kerur wrote:
>> > > On Sun, Apr 12, 2020 at 5:05 AM Tonghao Zhang <xiangxia.m.yue at gmail.com>
>> > > wrote:
>> > >
>> > >> On Sun, Apr 12, 2020 at 7:57 AM Ravi Kerur <rkerur at gmail.com> wrote:
>> > >>>
>> > >>> Hello OvS DPDK team,
>> > >>>
>> > >>> I am using OvS 2.13.90 and DPDK 19.11 and wanted to use HPET timer
>> > >> instead
>> > >>> of invariant TSC.
>> > >>>
>> > >>> Enabled CONFIG_RTE_LIBEAL_USE_HPET=y in DPDK
>> > >>>
>> > >>> Kernel has
>> > >>> CONFIG_HPET_TIMER=y
>> > >>> CONFIG_HPET_EMULATE_RTC=y
>> > >>> CONFIG_HPET=y
>> > >>> CONFIG_HPET_MMAP=y
>> > >>> CONFIG_HPET_MMAP_DEFAULT=y
>> > >>>
>> > >>> cat /sys/devices/system/clocksource/clocksource0/current_clocksource
>> > >>> tsc
>> > >>> # cat /sys/devices/system/clocksource/clocksource0/available_clocksource
>> > >>> tsc hpet acpi_pm
>> > >>>
>> > >>> OvS crashes at
>> > >>>
>> > >>> gdb ovsdb-server /core
>> > >>> GNU gdb (Ubuntu 8.1-0ubuntu3) 8.1.0.20180409-git
>> > >>> Copyright (C) 2018 Free Software Foundation, Inc.
>> > >>> License GPLv3+: GNU GPL version 3 or later <
>> > >> http://gnu.org/licenses/gpl.html
>> > >>>>
>> > >>> This is free software: you are free to change and redistribute it.
>> > >>> There is NO WARRANTY, to the extent permitted by law.  Type "show
>> > >> copying"
>> > >>> and "show warranty" for details.
>> > >>> This GDB was configured as "x86_64-linux-gnu".
>> > >>> Type "show configuration" for configuration details.
>> > >>> For bug reporting instructions, please see:
>> > >>> <http://www.gnu.org/software/gdb/bugs/>.
>> > >>> Find the GDB manual and other documentation resources online at:
>> > >>> <http://www.gnu.org/software/gdb/documentation/>.
>> > >>> For help, type "help".
>> > >>> Type "apropos word" to search for commands related to "word"...
>> > >>> Reading symbols from ovsdb-server...done.
>> > >>> [New LWP 9729]
>> > >>> [Thread debugging using libthread_db enabled]
>> > >>> Using host libthread_db library
>> > >> "/lib/x86_64-linux-gnu/libthread_db.so.1".
>> > >>> Core was generated by `ovsdb-server /usr/local/etc/openvswitch/conf.db
>> > >>> -vconsole:emer -vsyslog:err -vf'.
>> > >>> Program terminated with signal SIGSEGV, Segmentation fault.
>> > >>> #0  0x0000560fc6db8ac1 in rte_get_hpet_cycles ()
>> > >> Hi,
>> > >> This is a bug on dpdk, NOT ovs.
>> > >> I guess that your kernel does't support getentropy which supported in
>> > >> 3.17 and your cpu don't support rdseed.
>> > >> and that cause a dpdk bug. To workaround, you change: Not tested, hope
>> > >> it's useful for you
>> > >>
>> > >>
>> > > You are correct CPU doesn't support RDSEED instruction but RDRAND is
>> > > supported.
>> > >
>> > > It's a RT kernel based on 4.19 + RT patches.
>> > >
>> > > Setting CONFIG_RTE_LIBEAL_USE_HPET=n which is the default,'Invariant TSC'
>> > > will be used. However, I want to use HPET so should I check with DPDK dev?
>> >
>> > Yes.  This is a pure DPDK bug.  There is no way to fix it from the OVS side.
>> > Adding dpdk-dev in CC, moving ovs-dev to BCC to avoid cross posting.
>> Hi Ilya
>> I sent a patch to dpdk, but not applied. If applied, I feedback to this thread.
>> > Actually, looking at the code in DPDK, it's really old.  I'm wondering if it
>> > ever worked?  Looks like no-one ever tested this configuration.
>> >
>> > Best regards, Ilya Maximets.
>> >
>> > >
>> > > Thanks,
>> > > Ravi
>> > >
>> > >
>> > >
>> > >> diff --git a/lib/librte_eal/linux/eal_timer.c
>> > >> b/lib/librte_eal/linux/eal_timer.c
>> > >> index a904a8297cd2..4067e46da534 100644
>> > >> --- a/lib/librte_eal/linux/eal_timer.c
>> > >> +++ b/lib/librte_eal/linux/eal_timer.c
>> > >> @@ -26,7 +26,7 @@
>> > >>  #include "eal_private.h"
>> > >>  #include "eal_internal_cfg.h"
>> > >>
>> > >> -enum timer_source eal_timer_source = EAL_TIMER_HPET;
>> > >> +enum timer_source eal_timer_source = EAL_TIMER_TSC;
>> > >>
>> > >>  #ifdef RTE_LIBEAL_USE_HPET
>> > >>
>> > >>
>> > >>> (gdb) bt
>> > >>> #0  0x0000560fc6db8ac1 in rte_get_hpet_cycles ()
>> > >>> #1  0x0000560fc6c482b0 in rte_rand_init ()
>> > >>> #2  0x0000560fc6e125bd in __libc_csu_init ()
>> > >>> #3  0x00007f5e1f930b28 in __libc_start_main (main=0x560fc6c45940 <main>,
>> > >>> argc=14, argv=0x7fff61ae5088, init=0x560fc6e12570 <__libc_csu_init>,
>> > >>>     fini=<optimized out>, rtld_fini=<optimized out>,
>> > >>> stack_end=0x7fff61ae5078) at ../csu/libc-start.c:266
>> > >>> #4  0x0000560fc6c485da in _start ()
>> > >>> (gdb) quit
>> > >>>
>> > >>> RTE_INIT(rte_rand_init)
>> > >>> {
>> > >>> ...
>> > >>> }
>> > >>>
>> > >>> rte_rand_init is defined as initializer, following diffs won't help,
>> > >>>
>> > >>> diff --git a/lib/dpdk.c b/lib/dpdk.c
>> > >>> index 31450d470..f3d3989c8 100644
>> > >>> --- a/lib/dpdk.c
>> > >>> +++ b/lib/dpdk.c
>> > >>> @@ -405,6 +405,12 @@ dpdk_init__(const struct smap *ovs_other_config)
>> > >>>      if (result < 0) {
>> > >>>          VLOG_EMER("Unable to initialize DPDK: %s",
>> > >>> ovs_strerror(rte_errno));
>> > >>>          return false;
>> > >>> +    } else {
>> > >>> +#ifdef RTE_LIBEAL_USE_HPET
>> > >>> +        if (rte_eal_hpet_init(1) < 0) {
>> > >>> +            VLOG_ERR("HPET is not enabled, using TSC as default
>> > >> timer\n");
>> > >>> +        }
>> > >>> +#endif
>> > >>>      }
>> > >>>
>> > >>>      if (VLOG_IS_DBG_ENABLED()) {
>> > >>>
>> > >>> Kindly let me know your inputs on how to overcome this issue.
>> > >>>
>> > >>> Thanks,
>> > >>> Ravi
>> > >>> _______________________________________________
>> > >>> dev mailing list
>> > >>> dev at openvswitch.org
>> > >>> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>> > >>
>> > >>
>> > >>
>> > >> --
>> > >> Best regards, Tonghao
>> > >>
>> > > _______________________________________________
>> > > dev mailing list
>> > > dev at openvswitch.org
>> > > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>> > >
>> >
>>
>>
>> --
>> Best regards, Tonghao


-- 
Best regards, Tonghao