On 10/15/12 15:28, Gleb Natapov wrote: > On Mon, Oct 15, 2012 at 11:13:05AM +0200, Laszlo Ersek wrote: >> The "hang" is not actually a hang; the VM is active. I ftraced kvm in >> the epsilon-neighborhood of "point X", both under #6 (which hangs at X) >> and under SeaBIOS (which advances beyond X), and compared the traces. >> They are structurally similar (almost identical). Addresses seem to be >> "parallel", ie. offset by constant-looking values. Unfortunately I'm >> unable to identify the root cause from the few structural (and many >> address) differences. >> > Yes, Windows randomizes its address space which makes it harder to > debug. Can you provide example trace when VM "hangs".
http://people.redhat.com/~lersek/ovmf_win2k8r2sp1_debug/ftrace.tar.xz The command line was "trace-cmd record -b 20000 -e kvm". I lowered the number of VCPUs to 1 in the VM config. I made three attempts; the first two are about 2 seconds long each and "surround" the hang, while the third one is approx. 1 second long and was made firmly after the hang ensued. The third trace looks like a loop with the following body, iterations being cca. 15 msecs apart: qemu-kvm-1086 [001] 25169.909638: kvm_set_irq: gsi 8 level 1 source 0 qemu-kvm-1086 [001] 25169.909641: kvm_pic_set_irq: chip 1 pin 0 (edge|masked) qemu-kvm-1086 [001] 25169.909643: kvm_apic_accept_irq: apicid 0 vec 209 (Fixed|edge) qemu-kvm-1086 [001] 25169.909647: kvm_ioapic_set_irq: pin 8 dst 1 vec=209 (Fixed|logical|edge) qemu-kvm-1108 [002] 25169.909698: kvm_inj_virq: irq 209 qemu-kvm-1108 [002] 25169.909700: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909703: kvm_exit: reason EXCEPTION_NMI rip 0xfffff8000e4e1a6e qemu-kvm-1108 [002] 25169.909706: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909708: kvm_exit: reason IO_INSTRUCTION rip 0xfffff8000e4300b9 qemu-kvm-1108 [002] 25169.909709: kvm_pio: pio_write at 0x70 size 1 count 1 qemu-kvm-1108 [002] 25169.909724: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909725: kvm_exit: reason IO_INSTRUCTION rip 0xfffff8000e4300bf qemu-kvm-1108 [002] 25169.909726: kvm_pio: pio_read at 0x71 size 1 count 1 qemu-kvm-1108 [002] 25169.909731: kvm_set_irq: gsi 8 level 0 source 0 qemu-kvm-1108 [002] 25169.909732: kvm_pic_set_irq: chip 1 pin 0 (edge|masked) qemu-kvm-1108 [002] 25169.909732: kvm_ioapic_set_irq: pin 8 dst 1 vec=209 (Fixed|logical|edge) qemu-kvm-1108 [002] 25169.909736: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909737: kvm_exit: reason IO_INSTRUCTION rip 0xfffff8000e4300c0 qemu-kvm-1108 [002] 25169.909738: kvm_pio: pio_read at 0x71 size 1 count 1 qemu-kvm-1108 [002] 25169.909741: kvm_set_irq: gsi 8 level 0 source 0 qemu-kvm-1108 [002] 25169.909741: kvm_pic_set_irq: chip 1 pin 0 (edge|masked) qemu-kvm-1108 [002] 25169.909741: kvm_ioapic_set_irq: pin 8 dst 1 vec=209 (Fixed|logical|edge) qemu-kvm-1108 [002] 25169.909744: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909747: kvm_exit: reason APIC_ACCESS rip 0xfffff8000e439a70 qemu-kvm-1108 [002] 25169.909753: kvm_mmio: mmio write len 4 gpa 0xfee000b0 val 0x0 qemu-kvm-1108 [002] 25169.909753: kvm_apic: apic_write APIC_EOI = 0x0 qemu-kvm-1108 [002] 25169.909754: kvm_ack_irq: irqchip IOAPIC pin 8 qemu-kvm-1108 [002] 25169.909755: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909757: kvm_exit: reason EXCEPTION_NMI rip 0xfffff8000e4e1cfc qemu-kvm-1108 [002] 25169.909759: kvm_entry: vcpu 0 qemu-kvm-1108 [002] 25169.909769: kvm_exit: reason HLT rip 0xfffff8000e4397d1 with qemu-kvm-1108 [000] 25169.925098: kvm_entry: vcpu 0 qemu-kvm-1108 [000] 25169.925116: kvm_exit: reason EXTERNAL_INTERRUPT rip 0xfffff9600021c279 optionally / randomly mixed into some iterations (the rip varies). 0x70/0x71 seem to correspond to the CMOS RTC. 0xfee000b0 falls into the 1MB LAPIC range @ 0xFEE00000. Thank you, Gleb! Laszlo ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ edk2-devel mailing list edk2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/edk2-devel
