Nimrod Levy
2018-Feb-23 20:33 UTC
Ryzen lockup on bhyve was (Re: new Ryzen lockup issue ?)
After a couple of hours of running the iperf commands you were testing with, I'm unable to duplicate this so far. I'm running with FreeBSD stable from 17-Feb with the commits noted in https://reviews.freebsd.org/D14347 pulled in. I've also lowered the memory clock and disabled c-states in the bios. The bhyve VM is running CentOS. The system has been up for over 6 days and has been running the iperf3 loop for over 2 hours. The hardware is an Asus prime B350-Plus with a Ryzen 5 1600 and 32G of RAM. -- Nimrod On Fri, Feb 23, 2018 at 3:22 PM, Mike Tancsa <mike at sentex.net> wrote:> Actually I can confirm the same sort of hard lockup happens on my Epyc > board with RELENG11. It also happens in current. I will file a PR and > post on freebsd-current in case someone has any suggestions on how to > try and figure out whats going on. > > I upgraded the box to > 12.0-CURRENT #0 r329866 > in order to see if it could avoid the lockup, but same deal. The vmm > driver does seem different when loaded, but the same lock up under load > > CPU: AMD Ryzen 5 1600X Six-Core Processor (3593.35-MHz > K8-class CPU) > Origin="AuthenticAMD" Id=0x800f11 Family=0x17 Model=0x1 Stepping=1 > > Features=0x178bfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC, > SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,MMX,FXSR,SSE,SSE2,HTT> > > Features2=0x7ed8320b<SSE3,PCLMULQDQ,MON,SSSE3,FMA,CX16,SSE4. > 1,SSE4.2,MOVBE,POPCNT,AESNI,XSAVE,OSXSAVE,AVX,F16C,RDRAND> > AMD Features=0x2e500800<SYSCALL,NX,MMX+,FFXSR,Page1GB,RDTSCP,LM> > AMD > Features2=0x35c233ff<LAHF,CMP,SVM,ExtAPIC,CR8,ABM,SSE4A,MAS, > Prefetch,OSVW,SKINIT,WDT,TCE,Topology,PCXC,PNXC,DBE,PL2I,MWAITX> > Structured Extended > Features=0x209c01a9<FSGSBASE,BMI1,AVX2,SMEP,BMI2,RDSEED,ADX, > SMAP,CLFLUSHOPT,SHA> > XSAVE Features=0xf<XSAVEOPT,XSAVEC,XINUSE,XSAVES> > AMD Extended Feature Extensions ID EBX=0x7<CLZERO,IRPerf,XSaveErPtr> > SVM: NP,NRIP,VClean,AFlush,DAssist,NAsids=32768 > TSC: P-state invariant, performance statistics > > > AMD-Vi: IVRS Info VAsize = 64 PAsize = 48 GVAsize = 2 flags:0 > driver bug: Unable to set devclass (class: ppc devname: (unknown)) > ivhd0: <AMD-Vi/IOMMU ivhd with EFR> on acpi0 > ivhd0: Flag:b0<IotlbSup,Coherent> > ivhd0: Features(type:0x11) MsiNumPPR = 0 PNBanks= 2 PNCounters= 0 > ivhd0: Extended features[31:0]:22294ada<PPRSup,NXSup,GTSup,IASup> HATS > 0x2 GATS = 0x0 GLXSup = 0x1 SmiFSup = 0x1 SmiFRC = 0x2 GAMSup = 0x1 > DualPortLogSup = 0x2 DualEventLogSup = 0x2 > ivhd0: Extended features[62:32]:f77ef<USSup> Max PASID: 0x2f > DevTblSegSup = 0x3 MarcSup = 0x1 > ivhd0: supported paging level:7, will use only: 4 > ivhd0: device range: 0x0 - 0xffff > ivhd0: PCI cap 0x190b640f at 0x40 feature:19<IOTLB,EFR,CapExt> > > > > On 2/23/2018 12:35 PM, Nimrod Levy wrote: > > Now that is a fascinating data point. My machine that I've been having > > issues with has been running a bhyve vm from the beginning. I never > > made the connection. I'll try throwing some network traffic at the VM > > and see if I can make it lock up. > > > > On Fri, Feb 23, 2018 at 10:14 AM, Mike Tancsa <mike at sentex.net > > <mailto:mike at sentex.net>> wrote: > > > > On 2/22/2018 3:41 PM, Mike Tancsa wrote: > > > On 2/21/2018 3:04 PM, Mike Tancsa wrote: > > >> Not sure if I have found another issue specific to Ryzen, or a > bug that > > >> manifests itself on Ryzen systems easier. I installed the latest > > >> virtualbox from the ports and was doing some network performance > tests > > >> between a vm and the hypervisor using iperf3. The guest is just a > > >> RELENG11 image and the network is an em nic bridged to epair1b > > > > > > This looks possibly related to VirtualBox. Doing the same tests > and more > > > using bhyve, I dont get any lockup. Not to mention, network IO is > MUCH > > > faster. > > > > > > Actually, it just took a little bit longer to lock up the box with > bhyve > > on RELENG_11 as the hypervisor. Would be great if anyone can > confirm > > this locks up their Ryzen boxes ? I tried 2 different boxes to > eliminate > > a hardware issue. Also tried a similar test on Ubuntu and I can > spin up > > 4 instances and run without lockups. > > > > Just grab a copy of > > > > https://download.freebsd.org/ftp/releases/VM-IMAGES/11.1-RE > LEASE/amd64/Latest/FreeBSD-11.1-RELEASE-amd64.raw.xz > > <https://download.freebsd.org/ftp/releases/VM-IMAGES/11. > 1-RELEASE/amd64/Latest/FreeBSD-11.1-RELEASE-amd64.raw.xz> > > > > and make 2 copies. tmp.raw and tmp2.raw > > > > > > kldload vmm > > ifconfig tap0 create > > ifconfig tap1 create > > ifconfig tap1 up > > ifconfig tap0 up > > ifconfig bridge0 create addm tap0 addm tap1 > > ifconfig bridge0 192.168.99.1/24 <http://192.168.99.1/24> > > > > screen -d -m sh /usr/share/examples/bhyve/vmrun.sh -c 4 -m 6144M -t > tap0 > > -d tmp.raw BSD11a > > screen -d -m sh /usr/share/examples/bhyve/vmrun.sh -c 4 -m 6144M -t > tap1 > > -d tmp2.raw BSD11b > > > > Install netperf on the 2 vms and give the vtnet interface > > 192.168.99.2/24 <http://192.168.99.2/24> and 192.168.99.3/24 > > <http://192.168.99.3/24> > > > > In both VMs pkg install iperf3 and start it up as > > iperf -s > > > > In the hypervisor, > > iperf -t 10000 -R -c 192.168.99.2 > > iperf -t 10000 -c 192.168.99.3 > > > > > > the box locks up solid after 5-20 min. Same hardware with Ubuntu and > > virtual box and 4 instances work fine, no lockups after a day so not > > sure whats up but it seems to be something with the Ryzen CPU > running as > > a hypervisor or with some type of load :( > > > > Prior to lockup I had a stream of netstat -m writing to a file every > 5 > > seconds. The last entry was below. It doesnt seem to be leak. > > > > Thu Feb 22 17:14:28 EST 2018 > > 8694/10281/18975 mbufs in use (current/cache/total) > > 8225/5211/13436/2038424 mbuf clusters in use > (current/cache/total/max) > > 8225/5184 mbuf+clusters out of packet secondary zone in use > > (current/cache) > > 461/3747/4208/1019211 4k (page size) jumbo clusters in use > > (current/cache/total/max) > > 0/0/0/301988 9k jumbo clusters in use (current/cache/total/max) > > 0/0/0/169868 16k jumbo clusters in use (current/cache/total/max) > > 20467K/27980K/48447K bytes allocated to network (current/cache/total) > > 0/0/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters) > > 0/0/0 requests for mbufs delayed (mbufs/clusters/mbuf+clusters) > > 0/0/0 requests for jumbo clusters delayed (4k/9k/16k) > > 0/0/0 requests for jumbo clusters denied (4k/9k/16k) > > 0 sendfile syscalls > > 0 sendfile syscalls completed without I/O request > > 0 requests for I/O initiated by sendfile > > 0 pages read by sendfile as part of a request > > 0 pages were valid at time of a sendfile request > > 0 pages were requested for read ahead by applications > > 0 pages were read ahead by sendfile > > 0 times sendfile encountered an already busy page > > 0 requests for sfbufs denied > > 0 requests for sfbufs delayed > > > > > > > > ---Mike > > > > > > > > > > -- > > ------------------- > > Mike Tancsa, tel +1 519 651 3400 x203 <(519)%20651-3400> > > <tel:%2B1%20519%20651%203400%20x203> > > Sentex Communications, mike at sentex.net <mailto:mike at sentex.net> > > Providing Internet services since 1994 www.sentex.net > > <http://www.sentex.net> > > Cambridge, Ontario Canada > > _______________________________________________ > > freebsd-stable at freebsd.org <mailto:freebsd-stable at freebsd.org> > > mailing list > > https://lists.freebsd.org/mailman/listinfo/freebsd-stable > > <https://lists.freebsd.org/mailman/listinfo/freebsd-stable> > > To unsubscribe, send any mail to > > "freebsd-stable-unsubscribe at freebsd.org > > <mailto:freebsd-stable-unsubscribe at freebsd.org>" > > > > > > > -- > ------------------- > Mike Tancsa, tel +1 519 651 3400 x203 > Sentex Communications, mike at sentex.net > Providing Internet services since 1994 www.sentex.net > Cambridge, Ontario Canada >
Mike Tancsa
2018-Feb-23 20:44 UTC
Ryzen lockup on bhyve was (Re: new Ryzen lockup issue ?)
On 2/23/2018 3:33 PM, Nimrod Levy wrote:> After a couple of hours of running the iperf commands you were testing > with, I'm unable to duplicate this so far. > > I'm running with FreeBSD stable from 17-Feb with the commits noted > in?https://reviews.freebsd.org/D14347 > <https://reviews.freebsd.org/D14347>?pulled in. > > I've also lowered the memory clock and disabled c-states in the bios. > > The bhyve VM is running CentOS. > > The system has been up for over 6 days and has been running the iperf3 > loop for over 2 hours. > > The hardware is an Asus prime B350-Plus with a Ryzen 5 1600 and 32G of RAM.Hmmm, Interesting. Thank you for testing. I have a slightly different chipset (X370-PRO). Wouldnt think that would make a difference. And considering the Epyc also crashes-- although that takes a LOT longer, I am not sure whats going on. There is yet another BIOS update for this board. So trying with that. The vmm driver does report things a little different, so we will see ivhd0: <AMD-Vi/IOMMU ivhd with EFR> on acpi0 ivhd0: Flag:b0<IotlbSup,Coherent> ivhd0: Features(type:0x11) MsiNumPPR = 0 PNBanks= 2 PNCounters= 0 ivhd0: Extended features[31:0]:22294ada<PPRSup,NXSup,GTSup,IASup> HATS 0x2 GATS = 0x0 GLXSup = 0x1 SmiFSup = 0x1 SmiFRC = 0x2 GAMSup = 0x1 DualPortLogSup = 0x2 DualEventLogSup = 0x2 ivhd0: Extended features[62:32]:f77ef<USSup> Max PASID: 0x2f DevTblSegSup = 0x3 MarcSup = 0x1 ivhd0: supported paging level:7, will use only: 4 ivhd0: device range: 0x0 - 0xffff ivhd0: PCI cap 0x190b640f at 0x40 feature:19<IOTLB,EFR,CapExt> The load also matters. Try with 2 FreeBSD instances. I am not able to quickly crash with just one. ---Mike -- ------------------- Mike Tancsa, tel +1 519 651 3400 x203 Sentex Communications, mike at sentex.net Providing Internet services since 1994 www.sentex.net Cambridge, Ontario Canada
Nimrod Levy
2018-Mar-17 18:01 UTC
Ryzen lockup on bhyve was (Re: new Ryzen lockup issue ?)
Looks like I got almost 4 full weeks before it locked up this morning :( On Fri, Feb 23, 2018 at 3:33 PM Nimrod Levy <nimrodl at gmail.com> wrote:> After a couple of hours of running the iperf commands you were testing > with, I'm unable to duplicate this so far. > > I'm running with FreeBSD stable from 17-Feb with the commits noted in > https://reviews.freebsd.org/D14347 pulled in. > > I've also lowered the memory clock and disabled c-states in the bios. > > The bhyve VM is running CentOS. > > The system has been up for over 6 days and has been running the iperf3 > loop for over 2 hours. > > The hardware is an Asus prime B350-Plus with a Ryzen 5 1600 and 32G of RAM. > > -- > Nimrod > > > On Fri, Feb 23, 2018 at 3:22 PM, Mike Tancsa <mike at sentex.net> wrote: > >> Actually I can confirm the same sort of hard lockup happens on my Epyc >> board with RELENG11. It also happens in current. I will file a PR and >> post on freebsd-current in case someone has any suggestions on how to >> try and figure out whats going on. >> >> I upgraded the box to >> 12.0-CURRENT #0 r329866 >> in order to see if it could avoid the lockup, but same deal. The vmm >> driver does seem different when loaded, but the same lock up under load >> >> CPU: AMD Ryzen 5 1600X Six-Core Processor (3593.35-MHz >> K8-class CPU) >> Origin="AuthenticAMD" Id=0x800f11 Family=0x17 Model=0x1 Stepping=1 >> >> >> Features=0x178bfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,MMX,FXSR,SSE,SSE2,HTT> >> >> >> Features2=0x7ed8320b<SSE3,PCLMULQDQ,MON,SSSE3,FMA,CX16,SSE4.1,SSE4.2,MOVBE,POPCNT,AESNI,XSAVE,OSXSAVE,AVX,F16C,RDRAND> >> AMD Features=0x2e500800<SYSCALL,NX,MMX+,FFXSR,Page1GB,RDTSCP,LM> >> AMD >> >> Features2=0x35c233ff<LAHF,CMP,SVM,ExtAPIC,CR8,ABM,SSE4A,MAS,Prefetch,OSVW,SKINIT,WDT,TCE,Topology,PCXC,PNXC,DBE,PL2I,MWAITX> >> Structured Extended >> >> Features=0x209c01a9<FSGSBASE,BMI1,AVX2,SMEP,BMI2,RDSEED,ADX,SMAP,CLFLUSHOPT,SHA> >> XSAVE Features=0xf<XSAVEOPT,XSAVEC,XINUSE,XSAVES> >> AMD Extended Feature Extensions ID EBX=0x7<CLZERO,IRPerf,XSaveErPtr> >> SVM: NP,NRIP,VClean,AFlush,DAssist,NAsids=32768 >> TSC: P-state invariant, performance statistics >> >> >> AMD-Vi: IVRS Info VAsize = 64 PAsize = 48 GVAsize = 2 flags:0 >> driver bug: Unable to set devclass (class: ppc devname: (unknown)) >> ivhd0: <AMD-Vi/IOMMU ivhd with EFR> on acpi0 >> ivhd0: Flag:b0<IotlbSup,Coherent> >> ivhd0: Features(type:0x11) MsiNumPPR = 0 PNBanks= 2 PNCounters= 0 >> ivhd0: Extended features[31:0]:22294ada<PPRSup,NXSup,GTSup,IASup> HATS >> 0x2 GATS = 0x0 GLXSup = 0x1 SmiFSup = 0x1 SmiFRC = 0x2 GAMSup = 0x1 >> DualPortLogSup = 0x2 DualEventLogSup = 0x2 >> ivhd0: Extended features[62:32]:f77ef<USSup> Max PASID: 0x2f >> DevTblSegSup = 0x3 MarcSup = 0x1 >> ivhd0: supported paging level:7, will use only: 4 >> ivhd0: device range: 0x0 - 0xffff >> ivhd0: PCI cap 0x190b640f at 0x40 feature:19<IOTLB,EFR,CapExt> >> >> >> >> On 2/23/2018 12:35 PM, Nimrod Levy wrote: >> > Now that is a fascinating data point. My machine that I've been having >> > issues with has been running a bhyve vm from the beginning. I never >> > made the connection. I'll try throwing some network traffic at the VM >> > and see if I can make it lock up. >> > >> > On Fri, Feb 23, 2018 at 10:14 AM, Mike Tancsa <mike at sentex.net >> > <mailto:mike at sentex.net>> wrote: >> > >> > On 2/22/2018 3:41 PM, Mike Tancsa wrote: >> > > On 2/21/2018 3:04 PM, Mike Tancsa wrote: >> > >> Not sure if I have found another issue specific to Ryzen, or a >> bug that >> > >> manifests itself on Ryzen systems easier. I installed the latest >> > >> virtualbox from the ports and was doing some network performance >> tests >> > >> between a vm and the hypervisor using iperf3. The guest is just >> a >> > >> RELENG11 image and the network is an em nic bridged to epair1b >> > > >> > > This looks possibly related to VirtualBox. Doing the same tests >> and more >> > > using bhyve, I dont get any lockup. Not to mention, network IO >> is MUCH >> > > faster. >> > >> > >> > Actually, it just took a little bit longer to lock up the box with >> bhyve >> > on RELENG_11 as the hypervisor. Would be great if anyone can >> confirm >> > this locks up their Ryzen boxes ? I tried 2 different boxes to >> eliminate >> > a hardware issue. Also tried a similar test on Ubuntu and I can >> spin up >> > 4 instances and run without lockups. >> > >> > Just grab a copy of >> > >> > >> https://download.freebsd.org/ftp/releases/VM-IMAGES/11.1-RELEASE/amd64/Latest/FreeBSD-11.1-RELEASE-amd64.raw.xz >> > < >> https://download.freebsd.org/ftp/releases/VM-IMAGES/11.1-RELEASE/amd64/Latest/FreeBSD-11.1-RELEASE-amd64.raw.xz >> > >> > >> > and make 2 copies. tmp.raw and tmp2.raw >> > >> > >> > kldload vmm >> > ifconfig tap0 create >> > ifconfig tap1 create >> > ifconfig tap1 up >> > ifconfig tap0 up >> > ifconfig bridge0 create addm tap0 addm tap1 >> > ifconfig bridge0 192.168.99.1/24 <http://192.168.99.1/24> >> > >> > screen -d -m sh /usr/share/examples/bhyve/vmrun.sh -c 4 -m 6144M -t >> tap0 >> > -d tmp.raw BSD11a >> > screen -d -m sh /usr/share/examples/bhyve/vmrun.sh -c 4 -m 6144M -t >> tap1 >> > -d tmp2.raw BSD11b >> > >> > Install netperf on the 2 vms and give the vtnet interface >> > 192.168.99.2/24 <http://192.168.99.2/24> and 192.168.99.3/24 >> > <http://192.168.99.3/24> >> > >> > In both VMs pkg install iperf3 and start it up as >> > iperf -s >> > >> > In the hypervisor, >> > iperf -t 10000 -R -c 192.168.99.2 >> > iperf -t 10000 -c 192.168.99.3 >> > >> > >> > the box locks up solid after 5-20 min. Same hardware with Ubuntu >> and >> > virtual box and 4 instances work fine, no lockups after a day so not >> > sure whats up but it seems to be something with the Ryzen CPU >> running as >> > a hypervisor or with some type of load :( >> > >> > Prior to lockup I had a stream of netstat -m writing to a file >> every 5 >> > seconds. The last entry was below. It doesnt seem to be leak. >> > >> > Thu Feb 22 17:14:28 EST 2018 >> > 8694/10281/18975 mbufs in use (current/cache/total) >> > 8225/5211/13436/2038424 mbuf clusters in use >> (current/cache/total/max) >> > 8225/5184 mbuf+clusters out of packet secondary zone in use >> > (current/cache) >> > 461/3747/4208/1019211 4k (page size) jumbo clusters in use >> > (current/cache/total/max) >> > 0/0/0/301988 9k jumbo clusters in use (current/cache/total/max) >> > 0/0/0/169868 16k jumbo clusters in use (current/cache/total/max) >> > 20467K/27980K/48447K bytes allocated to network >> (current/cache/total) >> > 0/0/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters) >> > 0/0/0 requests for mbufs delayed (mbufs/clusters/mbuf+clusters) >> > 0/0/0 requests for jumbo clusters delayed (4k/9k/16k) >> > 0/0/0 requests for jumbo clusters denied (4k/9k/16k) >> > 0 sendfile syscalls >> > 0 sendfile syscalls completed without I/O request >> > 0 requests for I/O initiated by sendfile >> > 0 pages read by sendfile as part of a request >> > 0 pages were valid at time of a sendfile request >> > 0 pages were requested for read ahead by applications >> > 0 pages were read ahead by sendfile >> > 0 times sendfile encountered an already busy page >> > 0 requests for sfbufs denied >> > 0 requests for sfbufs delayed >> > >> > >> > >> > ---Mike >> > >> > >> > >> > >> > -- >> > ------------------- >> > Mike Tancsa, tel +1 519 651 3400 x203 <(519)%20651-3400> >> > <tel:%2B1%20519%20651%203400%20x203> >> > Sentex Communications, mike at sentex.net <mailto:mike at sentex.net> >> > Providing Internet services since 1994 www.sentex.net >> > <http://www.sentex.net> >> > Cambridge, Ontario Canada >> > _______________________________________________ >> > freebsd-stable at freebsd.org <mailto:freebsd-stable at freebsd.org> >> > mailing list >> > https://lists.freebsd.org/mailman/listinfo/freebsd-stable >> > <https://lists.freebsd.org/mailman/listinfo/freebsd-stable> >> > To unsubscribe, send any mail to >> > "freebsd-stable-unsubscribe at freebsd.org >> > <mailto:freebsd-stable-unsubscribe at freebsd.org>" >> > >> > >> >> >> -- >> ------------------- >> Mike Tancsa, tel +1 519 651 3400 x203 >> Sentex Communications, mike at sentex.net >> Providing Internet services since 1994 www.sentex.net >> Cambridge, Ontario Canada >> > > ---- Nimrod