- 30 May, 2018 3 commits
-
-
Paul Mackerras authored
This merges in the ppc-kvm topic branch of the powerpc repository to get some changes on which future patches will depend, in particular some new exports and TEXASR bit definitions. Signed-off-by:Paul Mackerras <paulus@ozlabs.org>
-
Paul Mackerras authored
This splits out the handling of "fake suspend" mode, part of the hypervisor TM assist code for POWER9, and puts almost all of it in new kvmppc_save_tm_hv and kvmppc_restore_tm_hv functions. The new functions branch to kvmppc_save/restore_tm if the CPU does not require hypervisor TM assistance. With this, it will be more straightforward to move kvmppc_save_tm and kvmppc_restore_tm to another file and use them for transactional memory support in PR KVM. Additionally, it also makes the code a bit clearer and reduces the number of feature sections. Signed-off-by: -
Paul Mackerras authored
Currently, PR KVM does not implement the configure_mmu operation, and so the KVM_PPC_CONFIGURE_V3_MMU ioctl always fails with an EINVAL error. This causes recent kernels to fail to boot as a PR KVM guest on POWER9, since recent kernels booted in HPT mode do the H_REGISTER_PROC_TBL hypercall, which causes userspace (QEMU) to do KVM_PPC_CONFIGURE_V3_MMU, which fails. This implements a minimal configure_mmu operation for PR KVM. It succeeds only if the MMU is being configured for HPT mode and no process table is being registered. This is enough to get recent kernels to boot as a PR KVM guest. Reviewed-by:
Greg Kurz <groug@kaod.org> Tested-by:
-
- 24 May, 2018 3 commits
-
-
Simon Guo authored
This patch exports tm_enable()/tm_disable/tm_abort() APIs, which will be used for PR KVM transactional memory logic. Signed-off-by:
Simon Guo <wei.guo.simon@gmail.com> Reviewed-by:
Michael Ellerman <mpe@ellerman.id.au>
-
Simon Guo authored
This patches add some macros for CR0/TEXASR bits so that PR KVM TM logic (tbegin./treclaim./tabort.) can make use of them later. Signed-off-by:
-
Simon Guo authored
PR KVM will need to reuse msr_check_and_set(). This patch exports this API for reuse. Signed-off-by:
-
- 22 May, 2018 7 commits
-
-
Simon Guo authored
This patch reimplements LOAD_VMX/STORE_VMX MMIO emulation with analyse_instr() input. When emulating the store, the VMX reg will need to be flushed so that the right reg val can be retrieved before writing to IO MEM. This patch also adds support for lvebx/lvehx/lvewx/stvebx/stvehx/stvewx MMIO emulation. To meet the requirement of handling different element sizes, kvmppc_handle_load128_by2x64()/kvmppc_handle_store128_by2x64() were replaced with kvmppc_handle_vmx_load()/kvmppc_handle_vmx_store(). The framework used is similar to VSX instruction MMIO emulation. Suggested-by:
-
Simon Guo authored
VSX MMIO emulation uses mmio_vsx_copy_type to represent VSX emulated element size/type, such as KVMPPC_VSX_COPY_DWORD_LOAD, etc. This patch expands mmio_vsx_copy_type to cover VMX copy type, such as KVMPPC_VMX_COPY_BYTE(stvebx/lvebx), etc. As a result, mmio_vsx_copy_type is also renamed to mmio_copy_type. It is a preparation for reimplementing VMX MMIO emulation. Signed-off-by:
-
Simon Guo authored
This patch reimplements LOAD_VSX/STORE_VSX instruction MMIO emulation with analyse_instr() input. It utilizes VSX_FPCONV/VSX_SPLAT/SIGNEXT exported by analyse_instr() and handle accordingly. When emulating VSX store, the VSX reg will need to be flushed so that the right reg val can be retrieved before writing to IO MEM. [paulus@ozlabs.org - mask the register number to 5 bits.] Suggested-by:
-
Simon Guo authored
This patch reimplements LOAD_FP/STORE_FP instruction MMIO emulation with analyse_instr() input. It utilizes the FPCONV/UPDATE properties exported by analyse_instr() and invokes kvmppc_handle_load(s)/kvmppc_handle_store() accordingly. For FP store MMIO emulation, the FP regs need to be flushed firstly so that the right FP reg vals can be read from vcpu->arch.fpr, which will be stored into MMIO data. Suggested-by:
-
Simon Guo authored
Currently HV will save math regs(FP/VEC/VSX) when trap into host. But PR KVM will only save math regs when qemu task switch out of CPU, or when returning from qemu code. To emulate FP/VEC/VSX mmio load, PR KVM need to make sure that math regs were flushed firstly and then be able to update saved VCPU FPR/VEC/VSX area reasonably. This patch adds giveup_ext() field to KVM ops. Only PR KVM has non-NULL giveup_ext() ops. kvmppc_complete_mmio_load() can invoke that hook (when not NULL) to flush math regs accordingly, before updating saved register vals. Math regs flush is also necessary for STORE, which will be covered in later patch within this patch series. Signed-off-by:
-
Simon Guo authored
This patch reimplements non-SIMD LOAD/STORE instruction MMIO emulation with analyse_instr() input. It utilizes the BYTEREV/UPDATE/SIGNEXT properties exported by analyse_instr() and invokes kvmppc_handle_load(s)/kvmppc_handle_store() accordingly. It also moves CACHEOP type handling into the skeleton. instruction_type within kvm_ppc.h is renamed to avoid conflict with sstep.h. Suggested-by:
-
Simon Guo authored
Some VSX instructions like lxvwsx will splat word into VSR. This patch adds a new VSX copy type KVMPPC_VSX_COPY_WORD_LOAD_DUMP to support this. Signed-off-by:
-
- 18 May, 2018 14 commits
-
-
Paul Mackerras authored
This relaxes the restriction on using PR KVM on POWER9. The existing code does work inside a guest partition running in HPT mode, because hypercalls such as H_ENTER use the old HPTE format, not the new format used by POWER9, and so no change to PR KVM's HPT manipulation code is required. PR KVM will still refuse to run if the kernel is using radix translation or if it is running bare-metal. Signed-off-by: -
Nicholas Piggin authored
It's possible to take a SRESET or MCE in these paths due to a bug in the host code or a NMI IPI, etc. A recent bug attempting to load a virtual address from real mode gave th complete but cryptic error, abridged: Oops: Bad interrupt in KVM entry/exit code, sig: 6 [#1] LE SMP NR_CPUS=2048 NUMA PowerNV CPU: 53 PID: 6582 Comm: qemu-system-ppc Not tainted NIP: c0000000000155ac LR: c0000000000c2430 CTR: c000000000015580 REGS: c000000fff76dd80 TRAP: 0200 Not tainted MSR: 9000000000201003 <SF,HV,ME,RI,LE> CR: 48082222 XER: 00000000 CFAR: 0000000102900ef0 DAR: d00017fffd941a28 DSISR: 00000040 SOFTE: 3 NIP [c0000000000155ac] perf_trace_tlbie+0x2c/0x1a0 LR [c0000000000c2430] do_tlbies+0x230/0x2f0 Sending the NMIs through the Linux handlers gives a nicer output: Severe Machine check interrupt [Not recovered] NIP [c0000000000155ac]: perf_trace_tlbie+0x2c/0x1a0 Initiator: CPU Error type: Real address [Load (bad)] Effective address: d00017fffcc01a28 opal: Machine check interrupt unrecoverable: MSR(RI=0) opal: Hardware platform error: Unrecoverable Machine Check exception CPU: 0 PID: 6700 Comm: qemu-system-ppc Tainted: G M NIP: c0000000000155ac LR: c0000000000c23c0 CTR: c000000000015580 REGS: c000000fff9e9d80 TRAP: 0200 Tainted: G M MSR: 9000000000201001 <SF,HV,ME,LE> CR: 48082222 XER: 00000000 CFAR: 000000010cbc1a30 DAR: d00017fffcc01a28 DSISR: 00000040 SOFTE: 3 NIP [c0000000000155ac] perf_trace_tlbie+0x2c/0x1a0 LR [c0000000000c23c0] do_tlbies+0x1c0/0x280 Signed-off-by:Nicholas Piggin <npiggin@gmail.com> Signed-off-by:
-
Nicholas Piggin authored
When CONFIG_RELOCATABLE=n, the Linux real mode interrupt handlers call into KVM using real address. This needs to be translated to the kernel linear effective address before the MMU is switched on. kvmppc_bad_host_intr misses adding these bits, so when it is used to handle a system reset interrupt (that always gets delivered in real mode), it results in an instruction access fault immediately after the MMU is turned on. Fix this by ensuring the top 2 address bits are set when the MMU is turned on. Signed-off-by:
-
Nicholas Piggin authored
Adding the write bit and RC bits to pte permissions does not require a pte clear and flush. There should not be other bits changed here, because restricting access or changing the PFN must have already invalidated any existing ptes (otherwise the race is already lost). Signed-off-by:
-
Nicholas Piggin authored
When the radix fault handler has no page from the process address space (e.g., for IO memory), it looks up the process pte and sets partition table pte using that to get attributes like CI and guarded. If the process table entry is to be writable, set _PAGE_DIRTY as well to avoid an RC update. If not, then ensure _PAGE_DIRTY does not come across. Set _PAGE_ACCESSED as well to avoid RC update. Signed-off-by:
-
Nicholas Piggin authored
The radix guest code can has fewer restrictions about what context it can run in, so move this flushing out of assembly and have it use the Linux TLB flush implementations introduced previously. This allows powerpc:tlbie trace events to be used. This changes the tlbiel sequence to only execute RIC=2 flush once on the first set flushed, then RIC=0 for the rest of the sets. The end result of the flush should be unchanged. This matches the local PID flush pattern that was introduced in a5998fcb ("powerpc/mm/radix: Optimise tlbiel flush all case"). Signed-off-by:
-
Nicholas Piggin authored
This has the advantage of consolidating TLB flush code in fewer places, and it also implements powerpc:tlbie trace events. Signed-off-by:
-
Nicholas Piggin authored
When partition scope mappings are unmapped with kvm_unmap_radix, the pte is cleared, but the page table structure is left in place. If the next page fault requests a different page table geometry (e.g., due to THP promotion or split), kvmppc_create_pte is responsible for changing the page tables. When a page table entry is to be converted to a large pte, the page table entry is cleared, the PWC flushed, then the page table it points to freed. This will cause pte page tables to leak when a 1GB page is to replace a pud entry points to a pmd table with pte tables under it: The pmd table will be freed, but its pte tables will be missed. Fix this by replacing the simple clear and free code with one that walks down the page tables and frees children. Care must be taken to clear the root entry being unmapped then flushing the PWC before freeing any page tables, as explained in comments. This requires PWC flush to logically become a flush-all-PWC (which it already is in hardware, but the KVM API needs to be changed to avoid confusion). This code also checks that no unexpected pte entries exist in any page table being freed, and unmaps those and emits a WARN. This is an expensive operation for the pte page level, but partition scope changes are rare, so it's unconditional for now to iron out bugs. It can be put under a CONFIG option or removed after some time. Signed-off-by:
-
Nicholas Piggin authored
Signed-off-by:
-
Nicholas Piggin authored
tlbies to an LPAR do not have to be serialised since POWER4/PPC970, after which the MMU_FTR_LOCKLESS_TLBIE feature was introduced to avoid tlbie locking. Since commit c17b98cf ("KVM: PPC: Book3S HV: Remove code for PPC970 processors"), KVM no longer supports processors that do not have this feature, so the tlbie locking can be removed completely. A sanity check for the feature is put in kvmppc_mmu_hv_init. Testing was done on a POWER9 system in HPT mode, with a -smp 32 guest in HPT mode. 32 instances of the powerpc fork benchmark from selftests were run with --fork, and the results measured. Without this patch, total throughput was about 13.5K/sec, and this is the top of the host profile: 74.52% [k] do_tlbies 2.95% [k] kvmppc_book3s_hv_page_fault 1.80% [k] calc_checksum 1.80% [k] kvmppc_vcpu_run_hv 1.49% [k] kvmppc_run_core After this patch, throughput was about 51K/sec, with this profile: 21.28% [k] do_tlbies 5.26% [k] kvmppc_run_core 4.88% [k] kvmppc_book3s_hv_page_fault 3.30% [k] _raw_spin_lock_irqsave 3.25% [k] gup_pgd_range Signed-off-by:
-
Simon Guo authored
When KVM emulates VMX store, it will invoke kvmppc_get_vmx_data() to retrieve VMX reg val. kvmppc_get_vmx_data() will check mmio_host_swabbed to decide which double word of vr[] to be used. But the mmio_host_swabbed can be uninitialized during VMX store procedure: kvmppc_emulate_loadstore \- kvmppc_handle_store128_by2x64 \- kvmppc_get_vmx_data So vcpu->arch.mmio_host_swabbed is not meant to be used at all for emulation of store instructions, and this patch makes that true for VMX stores. This patch also initializes mmio_host_swabbed to avoid possible future problems. Signed-off-by:
-
Simon Guo authored
This patch moves nip/ctr/lr/xer registers from scattered places in kvm_vcpu_arch to pt_regs structure. cr register is "unsigned long" in pt_regs and u32 in vcpu->arch. It will need more consideration and may move in later patches. Signed-off-by:
-
Simon Guo authored
Current regs are scattered at kvm_vcpu_arch structure and it will be more neat to organize them into pt_regs structure. Also it will enable reimplementation of MMIO emulation code with analyse_instr() later. Signed-off-by:
-
Paul Mackerras authored
This merges in the ppc-kvm topic branch of the powerpc repository to get some changes on which future patches will depend, in particular the definitions of various new TLB flushing functions. Signed-off-by:
-
- 17 May, 2018 13 commits
-
-
Souptick Joarder authored
Use new return type vm_fault_t for fault handler in struct vm_operations_struct. For now, this is just documenting that the function returns a VM_FAULT value rather than an errno. Once all instances are converted, vm_fault_t will become a distinct type. commit 1c8f4220 ("mm: change return type to vm_fault_t") Signed-off-by:
Souptick Joarder <jrdr.linux@gmail.com> Signed-off-by:
-
Alexey Kardashevskiy authored
Although it does not seem possible to break the host by passing bad parameters when creating a TCE table in KVM, it is still better to get an early clear indication of that than debugging weird effect this might bring. This adds some sanity checks that the page size is 4KB..16GB as this is what the actual LoPAPR supports and that the window actually fits 64bit space. Signed-off-by:
Alexey Kardashevskiy <aik@ozlabs.ru> Acked-by:
Balbir Singh <bsingharora@gmail.com> Signed-off-by:
-
Alexey Kardashevskiy authored
At the moment we only support in the host the IOMMU page sizes which the guest is aware of, which is 4KB/64KB/16MB. However P9 does not support 16MB IOMMU pages, 2MB and 1GB pages are supported instead. We can still emulate bigger guest pages (for example 16MB) with smaller host pages (4KB/64KB/2MB). This allows the physical IOMMU pages to use a page size smaller or equal than the guest visible IOMMU page size. Signed-off-by:
David Gibson <david@gibson.dropbear.id.au> Signed-off-by:
-
Alexey Kardashevskiy authored
The other TCE handlers use page shift from the guest visible TCE table (described by kvmppc_spapr_tce_iommu_table) so let's make H_STUFF_TCE handlers do the same thing. This should cause no behavioral change now but soon we will allow the iommu_table::it_page_shift being different from from the emulated table page size so this will play a role. Signed-off-by:
-
Paul Mackerras authored
We now have interrupts hard-disabled when coming back from kvmppc_hv_entry_trampoline, so this changes the comment to reflect that. Signed-off-by: -
Paul Mackerras authored
Although Linux doesn't use PURR and SPURR ((Scaled) Processor Utilization of Resources Register), other OSes depend on them. On POWER8 they count at a rate depending on whether the VCPU is idle or running, the activity of the VCPU, and the value in the RWMR (Region-Weighting Mode Register). Hardware expects the hypervisor to update the RWMR when a core is dispatched to reflect the number of online VCPUs in the vcore. This adds code to maintain a count in the vcore struct indicating how many VCPUs are online. In kvmppc_run_core we use that count to set the RWMR register on POWER8. If the core is split because of a static or dynamic micro-threading mode, we use the value for 8 threads. The RWMR value is not relevant when the host is executing because Linux does not use the PURR or SPURR register, so we don't bother saving and restoring the host value. For the sake of old userspace which does not set the KVM_REG_PPC_ONLINE register, we set online to 1 if it was 0 at the time of a KVM_RUN ioctl. Signed-off-by: -
Paul Mackerras authored
This adds a new KVM_REG_PPC_ONLINE register which userspace can set to 0 or 1 via the GET/SET_ONE_REG interface to indicate whether it considers the VCPU to be offline (0), that is, not currently running, or online (1). This will be used in a later patch to configure the register which controls PURR and SPURR accumulation. Signed-off-by: -
Paul Mackerras authored
A radix guest can execute tlbie instructions to invalidate TLB entries. After a tlbie or a group of tlbies, it must then do the architected sequence eieio; tlbsync; ptesync to ensure that the TLB invalidation has been processed by all CPUs in the system before it can rely on no CPU using any translation that it just invalidated. In fact it is the ptesync which does the actual synchronization in this sequence, and hardware has a requirement that the ptesync must be executed on the same CPU thread as the tlbies which it is expected to order. Thus, if a vCPU gets moved from one physical CPU to another after it has done some tlbies but before it can get to do the ptesync, the ptesync will not have the desired effect when it is executed on the second physical CPU. To fix this, we do a ptesync in the exit path for radix guests. If there are any pending tlbies, this will wait for them to complete. If there aren't, then ptesync will just do the same as sync. Signed-off-by: -
Benjamin Herrenschmidt authored
When a vcpu priority (CPPR) is set to a lower value (masking more interrupts), we stop processing interrupts already in the queue for the priorities that have now been masked. If those interrupts were previously re-routed to a different CPU, they might still be stuck until the older one that has them in its queue processes them. In the case of guest CPU unplug, that can be never. To address that without creating additional overhead for the normal interrupt processing path, this changes H_CPPR handling so that when such a priority change occurs, we scan the interrupt queue for that vCPU, and for any interrupt in there that has been re-routed, we replace it with a dummy and force a re-trigger. Signed-off-by:
Benjamin Herrenschmidt <benh@kernel.crashing.org> Tested-by:
-
Nicholas Piggin authored
The current partition table unmap code clears the _PAGE_PRESENT bit out of the pte, which leaves pud_huge/pmd_huge true and does not clear pud_present/pmd_present. This can confuse subsequent page faults and possibly lead to the guest looping doing continual hypervisor page faults. Signed-off-by:
-
Nicholas Piggin authored
The standard eieio ; tlbsync ; ptesync must follow tlbie to ensure it is ordered with respect to subsequent operations. Signed-off-by:
-
Paul Mackerras authored
Currently, the HV KVM guest entry/exit code adds the timebase offset from the vcore struct to the timebase on guest entry, and subtracts it on guest exit. Which is fine, except that it is possible for userspace to change the offset using the SET_ONE_REG interface while the vcore is running, as there is only one timebase offset per vcore but potentially multiple VCPUs in the vcore. If that were to happen, KVM would subtract a different offset on guest exit from that which it had added on guest entry, leading to the timebase being out of sync between cores in the host, which then leads to bad things happening such as hangs and spurious watchdog timeouts. To fix this, we add a new field 'tb_offset_applied' to the vcore struct which stores the offset that is currently applied to the timebase. This value is set from the vcore tb_offset field on guest entry, and is what is subtracted from the timebase on guest exit. Since it is zero when the timebase offset is not applied, we can simplify the logic in kvmhv_start_timing and kvmhv_accumulate_time. In addition, we had secondary threads reading the timebase while running concurrently with code on the primary thread which would eventually add or subtract the timebase offset from the timebase. This occurred while saving or restoring the DEC register value on the secondary threads. Although no specific incorrect behaviour has been observed, this is a race which should be fixed. To fix it, we move the DEC saving code to just before we call kvmhv_commence_exit, and the DEC restoring code to after the point where we have waited for the primary thread to switch the MMU context and add the timebase offset. That way we are sure that the timebase contains the guest timebase value in both cases. Signed-off-by: -
Mathieu Malaterre authored
Directly use fault_in_pages_readable instead of manual __get_user code. Fix warning treated as error with W=1: arch/powerpc/kernel/kvm.c:675:6: error: variable ‘tmp’ set but not used [-Werror=unused-but-set-variable] Suggested-by:
Christophe Leroy <christophe.leroy@c-s.fr> Signed-off-by:
Mathieu Malaterre <malat@debian.org> Reviewed-by:
-
