Re: [Qemu-devel] [Qemu-ppc] ppc/spapr: Radix guest causing host kernel oops

[joserz@linux.vnet.ibm.com]

On Mon, May 08, 2017 at 02:48:58PM +0530, Bharata B Rao wrote:
> Hi,
> 
> With ppc-for-2.10 branch of dwg's tree, starting a radix guest is currently
> causing a host kernel oops like this:
> 
> Unable to handle kernel paging request for data at address 0xe64bb17da64ab078
> Faulting instruction address: 0xc0000000002c3ddc
> Oops: Kernel access of bad area, sig: 11 [#1]
> SMP NR_CPUS=1024
> NUMA
> PowerNV
> task: c0000003bfb8b880 task.stack: c0000003c215c000
> NIP: c0000000002c3ddc LR: c0000000002c3e80 CTR: c0000000000ce2e0
> REGS: c0000003c215f150 TRAP: 0380   Not tainted  (4.11.0-1.git4a6869a.el7.centos.ppc64le)
> MSR: 9000000000001031 <SF,HV,ME,IR,DR,LE>
>   CR: 44008024  XER: 20000000
> CFAR: c0000000002c3e7c SOFTE: 1
> GPR00: 000000000000018f c0000003c215f3d0 c00000000131fd00 0000000000000000
> GPR04: 0000000000000005 00000000000001ff 0000000000000000 7db04aa67db14ba6
> GPR08: 264bb17da64ab000 e64bb17da64ab000 0000000000000078 0000000000000000
> GPR12: c0000003c32f0008 c00000000fdc0000 c00000000000e148 0000000000000000
> GPR16: 0000000008000000 0000000020000000 0000000000000000 c0000003c215f4c0
> GPR20: c0000001fd033000 c0000001fd0330e0 c0000001ffff8f50 c0000001f25afff8
> GPR24: 0000000000000200 00000001f25b0000 0000000000000010 0000000000020000
> GPR28: 0800000000000000 00000001f25b0000 000000007db04aa6 00000000a64ab07d
> NIP [c0000000002c3ddc] vmalloc_to_page+0x19c/0x220
> LR [c0000000002c3e80] vmalloc_to_pfn+0x20/0x50
> Call Trace:
> [c0000003c215f3d0] [7265677368657265] 0x7265677368657265 (unreliable)
> [c0000003c215f400] [c0000000002c3e80] vmalloc_to_pfn+0x20/0x50
> [c0000003c215f420] [c0000000000637e8] vmalloc_to_phys+0x28/0x60
> [c0000003c215f450] [c0000000000ce480] kvmppc_rm_h_put_tce_indirect+0x1a0/0x540
> [c0000003c215f590] [c0000000000d0314] hcall_try_real_mode+0x60/0x7c
> [c0000003c215f600] [c0000000000cefac] kvmppc_call_hv_entry+0x8/0x17c
> [c0000003c215f670] [c00800000357a970] __kvmppc_vcore_entry+0x13c/0x1ac [kvm_hv]
> [c0000003c215f840] [c0080000035774a8] kvmppc_run_core+0x788/0x1650 [kvm_hv]
> [c0000003c215fa00] [c0080000035790b8] kvmppc_vcpu_run_hv+0x388/0x1200 [kvm_hv]
> [c0000003c215fb30] [c008000002f34684] kvmppc_vcpu_run+0x34/0x50 [kvm]
> [c0000003c215fb50] [c008000002f30b54] kvm_arch_vcpu_ioctl_run+0x114/0x2a0 [kvm]
> [c0000003c215fbd0] [c008000002f23dd8] kvm_vcpu_ioctl+0x5e8/0x7c0 [kvm]
> [c0000003c215fd40] [c000000000350b50] do_vfs_ioctl+0xd0/0x8c0
> [c0000003c215fde0] [c000000000351414] SyS_ioctl+0xd4/0xf0
> [c0000003c215fe30] [c00000000000b8e0] system_call+0x38/0xfc
> Instruction dump:
> 53dfc42e 790807c6 394affff 7d08fb78 78638402 79081764 7d4a07b4 7c6a5038
> 7908f5e6 7d094b78 794a1f24 38600000 <7d2a482a> 7924cfe3 41820040 79260022
> 
> Reverting the below commit allows the radix guest to boot successfully.
> 
> commit 3dc410ae83e6cb76c81ea30a05d62596092b3165
> Author: Alexey Kardashevskiy <aik@ozlabs.ru>
> Date:   Mon Mar 27 16:22:19 2017 +1100
> 
>     target-ppc/kvm: Enable in-kernel TCE acceleration for multi-tce
>     
>     This enables in-kernel handling of H_PUT_TCE_INDIRECT and
>     H_STUFF_TCE hypercalls. The host kernel support is there since v4.6,
>     in particular d3695aa4f452
>     ("KVM: PPC: Add support for multiple-TCE hcalls").
>     
>     H_PUT_TCE is already accelerated and does not need any special enablement.
> 
> Regards,
> Bharata.
> 
> 

I'll send a revert commit, ok? At least to get some time to investigate
why this problem happens before re-enabling TCE again.

Thanks