- 30 Apr, 2019 30 commits
-
-
Paul Mackerras authored
This adds code to ensure that after a XIVE or XICS-on-XIVE KVM device is closed, KVM will not try to enable or disable any of the escalation interrupts for the VCPUs. We don't have to worry about races between clearing the pointers and use of the pointers by the XIVE context push/pull code, because the callers hold the vcpu->mutex, which is also taken by the KVM_RUN code. Therefore the vcpu cannot be entering or exiting the guest concurrently. Signed-off-by:Paul Mackerras <paulus@ozlabs.org>
-
Paul Mackerras authored
Now that we have the possibility of a XIVE or XICS-on-XIVE device being released while the VM is still running, we need to be careful about races and potential use-after-free bugs. Although the kvmppc_xive struct is not freed, but kept around for re-use, the kvmppc_xive_vcpu structs are freed, and they are used extensively in both the XIVE native and XICS-on-XIVE code. There are various ways in which XIVE code gets invoked: - VCPU entry and exit, which do push and pull operations on the XIVE hardware - one_reg get and set functions (vcpu->mutex is held) - XICS hypercalls (but only inside guest execution, not from kvmppc_pseries_do_hcall) - device creation calls (kvm->lock is held) - device callbacks - get/set attribute, mmap, pagefault, release/destroy - set_mapped/clr_mapped calls (kvm->lock is held) - connect_vcpu calls - debugfs file read callbacks Inside a device release function, we know that userspace cannot have an open file descriptor referring to the device, nor can it have any mmapped regions from the device. Therefore the device callbacks are excluded, as are the connect_vcpu calls (since they need a fd for the device). Further, since the caller holds the kvm->lock mutex, no other device creation calls or set/clr_mapped calls can be executing concurrently. To exclude VCPU execution and XICS hypercalls, we temporarily set kvm->arch.mmu_ready to 0. This forces any VCPU task that is trying to enter the guest to take the kvm->lock mutex, which is held by the caller of the release function. Then, sending an IPI to all other CPUs forces any VCPU currently executing in the guest to exit. Finally, we take the vcpu->mutex for each VCPU around the process of cleaning up and freeing its XIVE data structures, in order to exclude any one_reg get/set calls. To exclude the debugfs read callbacks, we just need to ensure that debugfs_remove is called before freeing any data structures. Once it returns we know that no CPU can be executing the callbacks (for our kvmppc_xive instance). Signed-off-by: -
Cédric Le Goater authored
When a P9 sPAPR VM boots, the CAS negotiation process determines which interrupt mode to use (XICS legacy or XIVE native) and invokes a machine reset to activate the chosen mode. We introduce 'release' methods for the XICS-on-XIVE and the XIVE native KVM devices which are called when the file descriptor of the device is closed after the TIMA and ESB pages have been unmapped. They perform the necessary cleanups : clear the vCPU interrupt presenters that could be attached and then destroy the device. The 'release' methods replace the 'destroy' methods as 'destroy' is not called anymore once 'release' is. Compatibility with older QEMU is nevertheless maintained. This is not considered as a safe operation as the vCPUs are still running and could be referencing the KVM device through their presenters. To protect the system from any breakage, the kvmppc_xive objects representing both KVM devices are now stored in an array under the VM. Allocation is performed on first usage and memory is freed only when the VM exits. [paulus@ozlabs.org - Moved freeing of xive structures to book3s.c, put it under #ifdef CONFIG_KVM_XICS.] Signed-off-by:
Cédric Le Goater <clg@kaod.org> Reviewed-by:
David Gibson <david@gibson.dropbear.id.au> Signed-off-by:
-
Cédric Le Goater authored
When a P9 sPAPR VM boots, the CAS negotiation process determines which interrupt mode to use (XICS legacy or XIVE native) and invokes a machine reset to activate the chosen mode. To be able to switch from one interrupt mode to another, we introduce the capability to release a KVM device without destroying the VM. The KVM device interface is extended with a new 'release' method which is called when the file descriptor of the device is closed. Once 'release' is called, the 'destroy' method will not be called anymore as the device is removed from the device list of the VM. Cc: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by:
-
Cédric Le Goater authored
Full support for the XIVE native exploitation mode is now available, advertise the capability KVM_CAP_PPC_IRQ_XIVE for guests running on PowerNV KVM Hypervisors only. Support for nested guests (pseries KVM Hypervisor) is not yet available. XIVE should also have been activated which is default setting on POWER9 systems running a recent Linux kernel. Signed-off-by:
-
Cédric Le Goater authored
The KVM XICS-over-XIVE device and the proposed KVM XIVE native device implement an IRQ space for the guest using the generic IPI interrupts of the XIVE IC controller. These interrupts are allocated at the OPAL level and "mapped" into the guest IRQ number space in the range 0-0x1FFF. Interrupt management is performed in the XIVE way: using loads and stores on the addresses of the XIVE IPI interrupt ESB pages. Both KVM devices share the same internal structure caching information on the interrupts, among which the xive_irq_data struct containing the addresses of the IPI ESB pages and an extra one in case of pass-through. The later contains the addresses of the ESB pages of the underlying HW controller interrupts, PHB4 in all cases for now. A guest, when running in the XICS legacy interrupt mode, lets the KVM XICS-over-XIVE device "handle" interrupt management, that is to perform the loads and stores on the addresses of the ESB pages of the guest interrupts. However, when running in XIVE native exploitation mode, the KVM XIVE native device exposes the interrupt ESB pages to the guest and lets the guest perform directly the loads and stores. The VMA exposing the ESB pages make use of a custom VM fault handler which role is to populate the VMA with appropriate pages. When a fault occurs, the guest IRQ number is deduced from the offset, and the ESB pages of associated XIVE IPI interrupt are inserted in the VMA (using the internal structure caching information on the interrupts). Supporting device passthrough in the guest running in XIVE native exploitation mode adds some extra refinements because the ESB pages of a different HW controller (PHB4) need to be exposed to the guest along with the initial IPI ESB pages of the XIVE IC controller. But the overall mechanic is the same. When the device HW irqs are mapped into or unmapped from the guest IRQ number space, the passthru_irq helpers, kvmppc_xive_set_mapped() and kvmppc_xive_clr_mapped(), are called to record or clear the passthrough interrupt information and to perform the switch. The approach taken by this patch is to clear the ESB pages of the guest IRQ number being mapped and let the VM fault handler repopulate. The handler will insert the ESB page corresponding to the HW interrupt of the device being passed-through or the initial IPI ESB page if the device is being removed. Signed-off-by:
-
Cédric Le Goater authored
Each source is associated with an Event State Buffer (ESB) with a even/odd pair of pages which provides commands to manage the source: to trigger, to EOI, to turn off the source for instance. The custom VM fault handler will deduce the guest IRQ number from the offset of the fault, and the ESB page of the associated XIVE interrupt will be inserted into the VMA using the internal structure caching information on the interrupts. Signed-off-by:
-
Cédric Le Goater authored
Each thread has an associated Thread Interrupt Management context composed of a set of registers. These registers let the thread handle priority management and interrupt acknowledgment. The most important are : - Interrupt Pending Buffer (IPB) - Current Processor Priority (CPPR) - Notification Source Register (NSR) They are exposed to software in four different pages each proposing a view with a different privilege. The first page is for the physical thread context and the second for the hypervisor. Only the third (operating system) and the fourth (user level) are exposed the guest. A custom VM fault handler will populate the VMA with the appropriate pages, which should only be the OS page for now. Signed-off-by: -
Cédric Le Goater authored
Some KVM devices will want to handle special mappings related to the underlying HW. For instance, the XIVE interrupt controller of the POWER9 processor has MMIO pages for thread interrupt management and for interrupt source control that need to be exposed to the guest when the OS has the required support. Cc: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by:
-
Cédric Le Goater authored
The state of the thread interrupt management registers needs to be collected for migration. These registers are cached under the 'xive_saved_state.w01' field of the VCPU when the VPCU context is pulled from the HW thread. An OPAL call retrieves the backup of the IPB register in the underlying XIVE NVT structure and merges it in the KVM state. Signed-off-by:
-
Cédric Le Goater authored
When migration of a VM is initiated, a first copy of the RAM is transferred to the destination before the VM is stopped, but there is no guarantee that the EQ pages in which the event notifications are queued have not been modified. To make sure migration will capture a consistent memory state, the XIVE device should perform a XIVE quiesce sequence to stop the flow of event notifications and stabilize the EQs. This is the purpose of the KVM_DEV_XIVE_EQ_SYNC control which will also marks the EQ pages dirty to force their transfer. Signed-off-by:
-
Cédric Le Goater authored
This control will be used by the H_INT_SYNC hcall from QEMU to flush event notifications on the XIVE IC owning the source. Signed-off-by:
-
Cédric Le Goater authored
This control is to be used by the H_INT_RESET hcall from QEMU. Its purpose is to clear all configuration of the sources and EQs. This is necessary in case of a kexec (for a kdump kernel for instance) to make sure that no remaining configuration is left from the previous boot setup so that the new kernel can start safely from a clean state. The queue 7 is ignored when the XIVE device is configured to run in single escalation mode. Prio 7 is used by escalations. The XIVE VP is kept enabled as the vCPU is still active and connected to the XIVE device. Signed-off-by:
-
Cédric Le Goater authored
These controls will be used by the H_INT_SET_QUEUE_CONFIG and H_INT_GET_QUEUE_CONFIG hcalls from QEMU to configure the underlying Event Queue in the XIVE IC. They will also be used to restore the configuration of the XIVE EQs and to capture the internal run-time state of the EQs. Both 'get' and 'set' rely on an OPAL call to access the EQ toggle bit and EQ index which are updated by the XIVE IC when event notifications are enqueued in the EQ. The value of the guest physical address of the event queue is saved in the XIVE internal xive_q structure for later use. That is when migration needs to mark the EQ pages dirty to capture a consistent memory state of the VM. To be noted that H_INT_SET_QUEUE_CONFIG does not require the extra OPAL call setting the EQ toggle bit and EQ index to configure the EQ, but restoring the EQ state will. Signed-off-by:
-
Cédric Le Goater authored
This control will be used by the H_INT_SET_SOURCE_CONFIG hcall from QEMU to configure the target of a source and also to restore the configuration of a source when migrating the VM. The XIVE source interrupt structure is extended with the value of the Effective Interrupt Source Number. The EISN is the interrupt number pushed in the event queue that the guest OS will use to dispatch events internally. Caching the EISN value in KVM eases the test when checking if a reconfiguration is indeed needed. Signed-off-by:
-
Cédric Le Goater authored
The XIVE KVM device maintains a list of interrupt sources for the VM which are allocated in the pool of generic interrupts (IPIs) of the main XIVE IC controller. These are used for the CPU IPIs as well as for virtual device interrupts. The IRQ number space is defined by QEMU. The XIVE device reuses the source structures of the XICS-on-XIVE device for the source blocks (2-level tree) and for the source interrupts. Under XIVE native, the source interrupt caches mostly configuration information and is less used than under the XICS-on-XIVE device in which hcalls are still necessary at run-time. When a source is initialized in KVM, an IPI interrupt source is simply allocated at the OPAL level and then MASKED. KVM only needs to know about its type: LSI or MSI. Signed-off-by:
-
Cédric Le Goater authored
The user interface exposes a new capability KVM_CAP_PPC_IRQ_XIVE to let QEMU connect the vCPU presenters to the XIVE KVM device if required. The capability is not advertised for now as the full support for the XIVE native exploitation mode is not yet available. When this is case, the capability will be advertised on PowerNV Hypervisors only. Nested guests (pseries KVM Hypervisor) are not supported. Internally, the interface to the new KVM device is protected with a new interrupt mode: KVMPPC_IRQ_XIVE. Signed-off-by:
-
Cédric Le Goater authored
This is the basic framework for the new KVM device supporting the XIVE native exploitation mode. The user interface exposes a new KVM device to be created by QEMU, only available when running on a L0 hypervisor. Support for nested guests is not available yet. The XIVE device reuses the device structure of the XICS-on-XIVE device as they have a lot in common. That could possibly change in the future if the need arise. Signed-off-by:
-
Paul Mackerras authored
This merges in the ppc-kvm topic branch from the powerpc tree to get patches which touch both general powerpc code and KVM code, one of which is a prerequisite for following patches. Signed-off-by: -
Suraj Jitindar Singh authored
On POWER9 and later processors where the host can schedule vcpus on a per thread basis, there is a streamlined entry path used when the guest is radix. This entry path saves/restores the fp and vr state in kvmhv_p9_guest_entry() by calling store_[fp/vr]_state() and load_[fp/vr]_state(). This is the same as the old entry path however the old entry path also saved/restored the VRSAVE register, which isn't done in the new entry path. This means that the vrsave register is now volatile across guest exit, which is an incorrect change in behaviour. Fix this by saving/restoring the vrsave register in kvmhv_p9_guest_entry(). This restores the old, correct, behaviour. Fixes: 95a6432c ("KVM: PPC: Book3S HV: Streamlined guest entry/exit path on P9 for radix guests") Signed-off-by:
Suraj Jitindar Singh <sjitindarsingh@gmail.com> Signed-off-by:
-
Paul Mackerras authored
When running on POWER9 with kvm_hv.indep_threads_mode = N and the host in SMT1 mode, KVM will run guest VCPUs on offline secondary threads. If those guests are in radix mode, we fail to load the LPID and flush the TLB if necessary, leading to the guest crashing with an unsupported MMU fault. This arises from commit 9a4506e1 ("KVM: PPC: Book3S HV: Make radix handle process scoped LPID flush in C, with relocation on", 2018-05-17), which didn't consider the case where indep_threads_mode = N. For simplicity, this makes the real-mode guest entry path flush the TLB in the same place for both radix and hash guests, as we did before 9a4506e1, though the code is now C code rather than assembly code. We also have the radix TLB flush open-coded rather than calling radix__local_flush_tlb_lpid_guest(), because the TLB flush can be called in real mode, and in real mode we don't want to invoke the tracepoint code. Fixes: 9a4506e1 ("KVM: PPC: Book3S HV: Make radix handle process scoped LPID flush in C, with relocation on") Signed-off-by:
-
Paul Mackerras authored
This replaces assembler code in book3s_hv_rmhandlers.S that checks the kvm->arch.need_tlb_flush cpumask and optionally does a TLB flush with C code in book3s_hv_builtin.c. Note that unlike the radix version, the hash version doesn't do an explicit ERAT invalidation because we will invalidate and load up the SLB before entering the guest, and that will invalidate the ERAT. Signed-off-by: -
Suraj Jitindar Singh authored
The code in book3s_hv_rmhandlers.S that pushes the XIVE virtual CPU context to the hardware currently assumes it is being called in real mode, which is usually true. There is however a path by which it can be executed in virtual mode, in the case where indep_threads_mode = N. A virtual CPU executing on an offline secondary thread can take a hypervisor interrupt in virtual mode and return from the kvmppc_hv_entry() call after the kvm_secondary_got_guest label. It is possible for it to be given another vcpu to execute before it gets to execute the stop instruction. In that case it will call kvmppc_hv_entry() for the second VCPU in virtual mode, and the XIVE vCPU push code will be executed in virtual mode. The result in that case will be a host crash due to an unexpected data storage interrupt caused by executing the stdcix instruction in virtual mode. This fixes it by adding a code path for virtual mode, which uses the virtual TIMA pointer and normal load/store instructions. [paulus@ozlabs.org - wrote patch description] Signed-off-by:
-
Paul Mackerras authored
This fixes a bug in the XICS emulation on POWER9 machines which is triggered by the guest doing a H_IPI with priority = 0 (the highest priority). What happens is that the notification interrupt arrives at the destination at priority zero. The loop in scan_interrupts() sees that a priority 0 interrupt is pending, but because xc->mfrr is zero, we break out of the loop before taking the notification interrupt out of the queue and EOI-ing it. (This doesn't happen when xc->mfrr != 0; in that case we process the priority-0 notification interrupt on the first iteration of the loop, and then break out of a subsequent iteration of the loop with hirq == XICS_IPI.) To fix this, we move the prio >= xc->mfrr check down to near the end of the loop. However, there are then some other things that need to be adjusted. Since we are potentially handling the notification interrupt and also delivering an IPI to the guest in the same loop iteration, we need to update pending and handle any q->pending_count value before the xc->mfrr check, rather than at the end of the loop. Also, we need to update the queue pointers when we have processed and EOI-ed the notification interrupt, since we may not do it later. Signed-off-by: -
Palmer Dabbelt authored
I made the same typo when trying to grep for uses of smp_wmb and figured I might as well fix it. Signed-off-by:
Palmer Dabbelt <palmer@sifive.com> Signed-off-by:
-
Alexey Kardashevskiy authored
We already allocate hardware TCE tables in multiple levels and skip intermediate levels when we can, now it is a turn of the KVM TCE tables. Thankfully these are allocated already in 2 levels. This moves the table's last level allocation from the creating helper to kvmppc_tce_put() and kvm_spapr_tce_fault(). Since such allocation cannot be done in real mode, this creates a virtual mode version of kvmppc_tce_put() which handles allocations. This adds kvmppc_rm_ioba_validate() to do an additional test if the consequent kvmppc_tce_put() needs a page which has not been allocated; if this is the case, we bail out to virtual mode handlers. The allocations are protected by a new mutex as kvm->lock is not suitable for the task because the fault handler is called with the mmap_sem held but kvmhv_setup_mmu() locks kvm->lock and mmap_sem in the reverse order. Signed-off-by:
Alexey Kardashevskiy <aik@ozlabs.ru> Signed-off-by:
-
Alexey Kardashevskiy authored
The kvmppc_tce_to_ua() helper is called from real and virtual modes and it works fine as long as CONFIG_DEBUG_LOCKDEP is not enabled. However if the lockdep debugging is on, the lockdep will most likely break in kvm_memslots() because of srcu_dereference_check() so we need to use PPC-own kvm_memslots_raw() which uses realmode safe rcu_dereference_raw_notrace(). This creates a realmode copy of kvmppc_tce_to_ua() which replaces kvm_memslots() with kvm_memslots_raw(). Since kvmppc_rm_tce_to_ua() becomes static and can only be used inside HV KVM, this moves it earlier under CONFIG_KVM_BOOK3S_HV_POSSIBLE. This moves truly virtual-mode kvmppc_tce_to_ua() to where it belongs and drops the prmap parameter which was never used in the virtual mode. Fixes: d3695aa4 ("KVM: PPC: Add support for multiple-TCE hcalls", 2016-02-15) Signed-off-by:
-
Alexey Kardashevskiy authored
The trace_hardirqs_on() sets current->hardirqs_enabled and from here the lockdep assumes interrupts are enabled although they are remain disabled until the context switches to the guest. Consequent srcu_read_lock() checks the flags in rcu_lock_acquire(), observes disabled interrupts and prints a warning (see below). This moves trace_hardirqs_on/off closer to __kvmppc_vcore_entry to prevent lockdep from being confused. DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled) WARNING: CPU: 16 PID: 8038 at kernel/locking/lockdep.c:4128 check_flags.part.25+0x224/0x280 [...] NIP [c000000000185b84] check_flags.part.25+0x224/0x280 LR [c000000000185b80] check_flags.part.25+0x220/0x280 Call Trace: [c000003fec253710] [c000000000185b80] check_flags.part.25+0x220/0x280 (unreliable) [c000003fec253780] [c000000000187ea4] lock_acquire+0x94/0x260 [c000003fec253840] [c00800001a1e9768] kvmppc_run_core+0xa60/0x1ab0 [kvm_hv] [c000003fec253a10] [c00800001a1ed944] kvmppc_vcpu_run_hv+0x73c/0xec0 [kvm_hv] [c000003fec253ae0] [c00800001a1095dc] kvmppc_vcpu_run+0x34/0x48 [kvm] [c000003fec253b00] [c00800001a1056bc] kvm_arch_vcpu_ioctl_run+0x2f4/0x400 [kvm] [c000003fec253b90] [c00800001a0f3618] kvm_vcpu_ioctl+0x460/0x850 [kvm] [c000003fec253d00] [c00000000041c4f4] do_vfs_ioctl+0xe4/0x930 [c000003fec253db0] [c00000000041ce04] ksys_ioctl+0xc4/0x110 [c000003fec253e00] [c00000000041ce78] sys_ioctl+0x28/0x80 [c000003fec253e20] [c00000000000b5a4] system_call+0x5c/0x70 Instruction dump: 419e0034 3d220004 39291730 81290000 2f890000 409e0020 3c82ffc6 3c62ffc5 3884be70 386329c0 4bf6ea71 60000000 <0fe00000> 3c62ffc6 3863be90 4801273d irq event stamp: 1025 hardirqs last enabled at (1025): [<c00800001a1e9728>] kvmppc_run_core+0xa20/0x1ab0 [kvm_hv] hardirqs last disabled at (1024): [<c00800001a1e9358>] kvmppc_run_core+0x650/0x1ab0 [kvm_hv] softirqs last enabled at (0): [<c0000000000f1210>] copy_process.isra.4.part.5+0x5f0/0x1d00 softirqs last disabled at (0): [<0000000000000000>] (null) ---[ end trace 31180adcc848993e ]--- possible reason: unannotated irqs-off. irq event stamp: 1025 hardirqs last enabled at (1025): [<c00800001a1e9728>] kvmppc_run_core+0xa20/0x1ab0 [kvm_hv] hardirqs last disabled at (1024): [<c00800001a1e9358>] kvmppc_run_core+0x650/0x1ab0 [kvm_hv] softirqs last enabled at (0): [<c0000000000f1210>] copy_process.isra.4.part.5+0x5f0/0x1d00 softirqs last disabled at (0): [<0000000000000000>] (null) Fixes: 8b24e69f ("KVM: PPC: Book3S HV: Close race with testing for signals on guest entry", 2017-06-26) Signed-off-by:
-
Suraj Jitindar Singh authored
Implement a real mode handler for the H_CALL H_PAGE_INIT which can be used to zero or copy a guest page. The page is defined to be 4k and must be 4k aligned. The in-kernel real mode handler halves the time to handle this H_CALL compared to handling it in userspace for a hash guest. Signed-off-by:
-
Suraj Jitindar Singh authored
Implement a virtual mode handler for the H_CALL H_PAGE_INIT which can be used to zero or copy a guest page. The page is defined to be 4k and must be 4k aligned. The in-kernel handler halves the time to handle this H_CALL compared to handling it in userspace for a radix guest. Signed-off-by:
-
- 20 Apr, 2019 1 commit
-
-
Michael Neuling authored
This adds a flag so that the DAWR can be enabled on P9 via: echo Y > /sys/kernel/debug/powerpc/dawr_enable_dangerous The DAWR was previously force disabled on POWER9 in: 96541531 powerpc: Disable DAWR in the base POWER9 CPU features Also see Documentation/powerpc/DAWR-POWER9.txt This is a dangerous setting, USE AT YOUR OWN RISK. Some users may not care about a bad user crashing their box (ie. single user/desktop systems) and really want the DAWR. This allows them to force enable DAWR. This flag can also be used to disable DAWR access. Once this is cleared, all DAWR access should be cleared immediately and your machine once again safe from crashing. Userspace may get confused by toggling this. If DAWR is force enabled/disabled between getting the number of breakpoints (via PTRACE_GETHWDBGINFO) and setting the breakpoint, userspace will get an inconsistent view of what's available. Similarly for guests. For the DAWR to be enabled in a KVM guest, the DAWR needs to be force enabled in the host AND the guest. For this reason, this won't work on POWERVM as it doesn't allow the HCALL to work. Writes of 'Y' to the dawr_enable_dangerous file will fail if the hypervisor doesn't support writing the DAWR. To double check the DAWR is working, run this kernel selftest: tools/testing/selftests/powerpc/ptrace/ptrace-hwbreak.c Any errors/failures/skips mean something is wrong. Signed-off-by:
Michael Neuling <mikey@neuling.org> Signed-off-by:
Michael Ellerman <mpe@ellerman.id.au>
-
- 11 Apr, 2019 1 commit
-
-
Cédric Le Goater authored
The support for XIVE native exploitation mode in Linux/KVM needs a couple more OPAL calls to get and set the state of the XIVE internal structures being used by a sPAPR guest. Signed-off-by:
-
- 05 Apr, 2019 2 commits
-
-
Alexey Kardashevskiy authored
Guest physical to user address translation uses KVM memslots and reading these requires holding the kvm->srcu lock. However recently introduced kvmppc_tce_validate() broke the rule (see the lockdep warning below). This moves srcu_read_lock(&vcpu->kvm->srcu) earlier to protect kvmppc_tce_validate() as well. ============================= WARNING: suspicious RCU usage 5.1.0-rc2-le_nv2_aikATfstn1-p1 #380 Not tainted ----------------------------- include/linux/kvm_host.h:605 suspicious rcu_dereference_check() usage! other info that might help us debug this: rcu_scheduler_active = 2, debug_locks = 1 1 lock held by qemu-system-ppc/8020: #0: 0000000094972fe9 (&vcpu->mutex){+.+.}, at: kvm_vcpu_ioctl+0xdc/0x850 [kvm] stack backtrace: CPU: 44 PID: 8020 Comm: qemu-system-ppc Not tainted 5.1.0-rc2-le_nv2_aikATfstn1-p1 #380 Call Trace: [c000003fece8f740] [c000000000bcc134] dump_stack+0xe8/0x164 (unreliable) [c000003fece8f790] [c000000000181be0] lockdep_rcu_suspicious+0x130/0x170 [c000003fece8f810] [c0000000000d5f50] kvmppc_tce_to_ua+0x280/0x290 [c000003fece8f870] [c00800001a7e2c78] kvmppc_tce_validate+0x80/0x1b0 [kvm] [c000003fece8f8e0] [c00800001a7e3fac] kvmppc_h_put_tce+0x94/0x3e4 [kvm] [c000003fece8f9a0] [c00800001a8baac4] kvmppc_pseries_do_hcall+0x30c/0xce0 [kvm_hv] [c000003fece8fa10] [c00800001a8bd89c] kvmppc_vcpu_run_hv+0x694/0xec0 [kvm_hv] [c000003fece8fae0] [c00800001a7d95dc] kvmppc_vcpu_run+0x34/0x48 [kvm] [c000003fece8fb00] [c00800001a7d56bc] kvm_arch_vcpu_ioctl_run+0x2f4/0x400 [kvm] [c000003fece8fb90] [c00800001a7c3618] kvm_vcpu_ioctl+0x460/0x850 [kvm] [c000003fece8fd00] [c00000000041c4f4] do_vfs_ioctl+0xe4/0x930 [c000003fece8fdb0] [c00000000041ce04] ksys_ioctl+0xc4/0x110 [c000003fece8fe00] [c00000000041ce78] sys_ioctl+0x28/0x80 [c000003fece8fe20] [c00000000000b5a4] system_call+0x5c/0x70 Fixes: 42de7b9e ("KVM: PPC: Validate TCEs against preregistered memory page sizes", 2018-09-10) Signed-off-by: -
Suraj Jitindar Singh authored
There is a hardware bug in some POWER9 processors where a treclaim in fake suspend mode can cause an inconsistency in the XER[SO] bit across the threads of a core, the workaround being to force the core into SMT4 when doing the treclaim. The FAKE_SUSPEND bit (bit 10) in the PSSCR is used to control whether a thread is in fake suspend or real suspend. The important difference here being that thread reconfiguration is blocked in real suspend but not fake suspend mode. When we exit a guest which was in fake suspend mode, we force the core into SMT4 while we do the treclaim in kvmppc_save_tm_hv(). However on the new exit path introduced with the function kvmhv_run_single_vcpu() we restore the host PSSCR before calling kvmppc_save_tm_hv() which means that if we were in fake suspend mode we put the thread into real suspend mode when we clear the PSSCR[FAKE_SUSPEND] bit. This means that we block thread reconfiguration and the thread which is trying to get the core into SMT4 before it can do the treclaim spins forever since it itself is blocking thread reconfiguration. The result is that that core is essentially lost. This results in a trace such as: [ 93.512904] CPU: 7 PID: 13352 Comm: qemu-system-ppc Not tainted 5.0.0 #4 [ 93.512905] NIP: c000000000098a04 LR: c0000000000cc59c CTR: 0000000000000000 [ 93.512908] REGS: c000003fffd2bd70 TRAP: 0100 Not tainted (5.0.0) [ 93.512908] MSR: 9000000302883033 <SF,HV,VEC,VSX,FP,ME,IR,DR,RI,LE,TM[SE]> CR: 22222444 XER: 00000000 [ 93.512914] CFAR: c000000000098a5c IRQMASK: 3 [ 93.512915] PACATMSCRATCH: 0000000000000001 [ 93.512916] GPR00: 0000000000000001 c000003f6cc1b830 c000000001033100 0000000000000004 [ 93.512928] GPR04: 0000000000000004 0000000000000002 0000000000000004 0000000000000007 [ 93.512930] GPR08: 0000000000000000 0000000000000004 0000000000000000 0000000000000004 [ 93.512932] GPR12: c000203fff7fc000 c000003fffff9500 0000000000000000 0000000000000000 [ 93.512935] GPR16: 2000000000300375 000000000000059f 0000000000000000 0000000000000000 [ 93.512951] GPR20: 0000000000000000 0000000000080053 004000000256f41f c000003f6aa88ef0 [ 93.512953] GPR24: c000003f6aa89100 0000000000000010 0000000000000000 0000000000000000 [ 93.512956] GPR28: c000003f9e9a0800 0000000000000000 0000000000000001 c000203fff7fc000 [ 93.512959] NIP [c000000000098a04] pnv_power9_force_smt4_catch+0x1b4/0x2c0 [ 93.512960] LR [c0000000000cc59c] kvmppc_save_tm_hv+0x40/0x88 [ 93.512960] Call Trace: [ 93.512961] [c000003f6cc1b830] [0000000000080053] 0x80053 (unreliable) [ 93.512965] [c000003f6cc1b8a0] [c00800001e9cb030] kvmhv_p9_guest_entry+0x508/0x6b0 [kvm_hv] [ 93.512967] [c000003f6cc1b940] [c00800001e9cba44] kvmhv_run_single_vcpu+0x2dc/0xb90 [kvm_hv] [ 93.512968] [c000003f6cc1ba10] [c00800001e9cc948] kvmppc_vcpu_run_hv+0x650/0xb90 [kvm_hv] [ 93.512969] [c000003f6cc1bae0] [c00800001e8f620c] kvmppc_vcpu_run+0x34/0x48 [kvm] [ 93.512971] [c000003f6cc1bb00] [c00800001e8f2d4c] kvm_arch_vcpu_ioctl_run+0x2f4/0x400 [kvm] [ 93.512972] [c000003f6cc1bb90] [c00800001e8e3918] kvm_vcpu_ioctl+0x460/0x7d0 [kvm] [ 93.512974] [c000003f6cc1bd00] [c0000000003ae2c0] do_vfs_ioctl+0xe0/0x8e0 [ 93.512975] [c000003f6cc1bdb0] [c0000000003aeb24] ksys_ioctl+0x64/0xe0 [ 93.512978] [c000003f6cc1be00] [c0000000003aebc8] sys_ioctl+0x28/0x80 [ 93.512981] [c000003f6cc1be20] [c00000000000b3a4] system_call+0x5c/0x70 [ 93.512983] Instruction dump: [ 93.512986] 419dffbc e98c0000 2e8b0000 38000001 60000000 60000000 60000000 40950068 [ 93.512993] 392bffff 39400000 79290020 39290001 <7d2903a6> 60000000 60000000 7d235214 To fix this we preserve the PSSCR[FAKE_SUSPEND] bit until we call kvmppc_save_tm_hv() which will mean the core can get into SMT4 and perform the treclaim. Note kvmppc_save_tm_hv() clears the PSSCR[FAKE_SUSPEND] bit again so there is no need to explicitly do that. Fixes: 95a6432c ("KVM: PPC: Book3S HV: Streamlined guest entry/exit path on P9 for radix guests") Signed-off-by:
-
- 28 Mar, 2019 6 commits
-
-
Paolo Bonzini authored
Merge tag 'kvmarm-fixes-for-5.1' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into kvm-master KVM/ARM fixes for 5.1 - Fix THP handling in the presence of pre-existing PTEs - Honor request for PTE mappings even when THPs are available - GICv4 performance improvement - Take the srcu lock when writing to guest-controlled ITS data structures - Reset the virtual PMU in preemptible context - Various cleanups
-
Paolo Bonzini authored
The documentation does not mention how to delete a slot, add the information. Reported-by:
Nathaniel McCallum <npmccallum@redhat.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com>
-
Sean Christopherson authored
The series to add memcg accounting to KVM allocations[1] states: There are many KVM kernel memory allocations which are tied to the life of the VM process and should be charged to the VM process's cgroup. While it is correct to account KVM kernel allocations to the cgroup of the process that created the VM, it's technically incorrect to state that the KVM kernel memory allocations are tied to the life of the VM process. This is because the VM itself, i.e. struct kvm, is not tied to the life of the process which created it, rather it is tied to the life of its associated file descriptor. In other words, kvm_destroy_vm() is not invoked until fput() decrements its associated file's refcount to zero. A simple example is to fork() in Qemu and have the child sleep indefinitely; kvm_destroy_vm() isn't called until Qemu closes its file descriptor *and* the rogue child is killed. The allocations are guaranteed to be *accounted* to the process which created the VM, but only because KVM's per-{VM,vCPU} ioctls reject the ioctl() with -EIO if kvm->mm != current->mm. I.e. the child can keep the VM "alive" but can't do anything useful with its reference. Note that because 'struct kvm' also holds a reference to the mm_struct of its owner, the above behavior also applies to userspace allocations. Given that mucking with a VM's file descriptor can lead to subtle and undesirable behavior, e.g. memcg charges persisting after a VM is shut down, explicitly document a VM's lifecycle and its impact on the VM's resources. Alternatively, KVM could aggressively free resources when the creating process exits, e.g. via mmu_notifier->release(). However, mmu_notifier isn't guaranteed to be available, and freeing resources when the creator exits is likely to be error prone and fragile as KVM would need to ensure that it only freed resources that are truly out of reach. In practice, the existing behavior shouldn't be problematic as a properly configured system will prevent a child process from being moved out of the appropriate cgroup hierarchy, i.e. prevent hiding the process from the OOM killer, and will prevent an unprivileged user from being able to to hold a reference to struct kvm via another method, e.g. debugfs. [1]https://patchwork.kernel.org/patch/10806707/Signed-off-by:Sean Christopherson <sean.j.christopherson@intel.com> Signed-off-by:
-
Sean Christopherson authored
Documentation/virtual/kvm/api.txt states: NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR and KVM_EXIT_EPR the corresponding operations are complete (and guest state is consistent) only after userspace has re-entered the kernel with KVM_RUN. The kernel side will first finish incomplete operations and then check for pending signals. Userspace can re-enter the guest with an unmasked signal pending to complete pending operations. Because guest state may be inconsistent, starting state migration after an IO exit without first completing IO may result in test failures, e.g. a proposed change to KVM's handling of %rip in its fast PIO handling[1] will cause the new VM, i.e. the post-migration VM, to have its %rip set to the IN instruction that triggered KVM_EXIT_IO, leading to a test assertion due to a stage mismatch. For simplicitly, require KVM_CAP_IMMEDIATE_EXIT to complete IO and skip the test if it's not available. The addition of KVM_CAP_IMMEDIATE_EXIT predates the state selftest by more than a year. [1] https://patchwork.kernel.org/patch/10848545/ Fixes: fa3899ad ("kvm: selftests: add basic test for state save and restore") Reported-by:Jim Mattson <jmattson@google.com> Signed-off-by:
-
Sean Christopherson authored
Since 4.8.3, gcc has enabled -fstack-protector by default. This is problematic for the KVM selftests as they do not configure fs or gs segments (the stack canary is pulled from fs:0x28). With the default behavior, gcc will insert a stack canary on any function that creates buffers of 8 bytes or more. As a result, ucall() will hit a triple fault shutdown due to reading a bad fs segment when inserting its stack canary, i.e. every test fails with an unexpected SHUTDOWN. Fixes: 14c47b75 ("kvm: selftests: introduce ucall") Signed-off-by:
-
Sean Christopherson authored
KVM selftests embed the guest "image" as a function in the test itself and extract the guest code at runtime by manually parsing the elf headers. The parsing is very simple and doesn't supporting fancy things like position independent executables. Recent versions of gcc enable pie by default, which results in triple fault shutdowns in the guest due to the virtual address in the headers not matching up with the virtual address retrieved from the function pointer. Signed-off-by:
-
