/*
* Interrupt code generation macros
*/
-#define IVEC .L_IVEC_\name\()
-#define IHSRR .L_IHSRR_\name\()
-#define IHSRR_IF_HVMODE .L_IHSRR_IF_HVMODE_\name\()
-#define IAREA .L_IAREA_\name\()
-#define IVIRT .L_IVIRT_\name\()
-#define IISIDE .L_IISIDE_\name\()
-#define IDAR .L_IDAR_\name\()
-#define IDSISR .L_IDSISR_\name\()
-#define ISET_RI .L_ISET_RI_\name\()
-#define IBRANCH_TO_COMMON .L_IBRANCH_TO_COMMON_\name\()
-#define IREALMODE_COMMON .L_IREALMODE_COMMON_\name\()
-#define IMASK .L_IMASK_\name\()
-#define IKVM_SKIP .L_IKVM_SKIP_\name\()
-#define IKVM_REAL .L_IKVM_REAL_\name\()
+#define IVEC .L_IVEC_\name\() /* Interrupt vector address */
+#define IHSRR .L_IHSRR_\name\() /* Sets SRR or HSRR registers */
+#define IHSRR_IF_HVMODE .L_IHSRR_IF_HVMODE_\name\() /* HSRR if HV else SRR */
+#define IAREA .L_IAREA_\name\() /* PACA save area */
+#define IVIRT .L_IVIRT_\name\() /* Has virt mode entry point */
+#define IISIDE .L_IISIDE_\name\() /* Uses SRR0/1 not DAR/DSISR */
+#define IDAR .L_IDAR_\name\() /* Uses DAR (or SRR0) */
+#define IDSISR .L_IDSISR_\name\() /* Uses DSISR (or SRR1) */
+#define ISET_RI .L_ISET_RI_\name\() /* Run common code w/ MSR[RI]=1 */
+#define IBRANCH_TO_COMMON .L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
+#define IREALMODE_COMMON .L_IREALMODE_COMMON_\name\() /* Common runs in realmode */
+#define IMASK .L_IMASK_\name\() /* IRQ soft-mask bit */
+#define IKVM_SKIP .L_IKVM_SKIP_\name\() /* Generate KVM skip handler */
+#define IKVM_REAL .L_IKVM_REAL_\name\() /* Real entry tests KVM */
#define __IKVM_REAL(name) .L_IKVM_REAL_ ## name
-#define IKVM_VIRT .L_IKVM_VIRT_\name\()
-#define ISTACK .L_ISTACK_\name\()
+#define IKVM_VIRT .L_IKVM_VIRT_\name\() /* Virt entry tests KVM */
+#define ISTACK .L_ISTACK_\name\() /* Set regular kernel stack */
#define __ISTACK(name) .L_ISTACK_ ## name
-#define IRECONCILE .L_IRECONCILE_\name\()
-#define IKUAP .L_IKUAP_\name\()
+#define IRECONCILE .L_IRECONCILE_\name\() /* Do RECONCILE_IRQ_STATE */
+#define IKUAP .L_IKUAP_\name\() /* Do KUAP lock */
#define INT_DEFINE_BEGIN(n) \
.macro int_define_ ## n name
EXC_VIRT_NONE(0x4000, 0x100)
+/**
+ * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
+ * This is a non-maskable, asynchronous interrupt always taken in real-mode.
+ * It is caused by:
+ * - Wake from power-saving state, on powernv.
+ * - An NMI from another CPU, triggered by firmware or hypercall.
+ * - As crash/debug signal injected from BMC, firmware or hypervisor.
+ *
+ * Handling:
+ * Power-save wakeup is the only performance critical path, so this is
+ * determined quickly as possible first. In this case volatile registers
+ * can be discarded and SPRs like CFAR don't need to be read.
+ *
+ * If not a powersave wakeup, then it's run as a regular interrupt, however
+ * it uses its own stack and PACA save area to preserve the regular kernel
+ * environment for debugging.
+ *
+ * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
+ * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
+ * correct to switch to virtual mode to run the regular interrupt handler
+ * because it might be interrupted when the MMU is in a bad state (e.g., SLB
+ * is clear).
+ *
+ * FWNMI:
+ * PAPR specifies a "fwnmi" facility which sends the sreset to a different
+ * entry point with a different register set up. Some hypervisors will
+ * send the sreset to 0x100 in the guest if it is not fwnmi capable.
+ *
+ * KVM:
+ * Unlike most SRR interrupts, this may be taken by the host while executing
+ * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
+ * mode and then raise the sreset.
+ */
INT_DEFINE_BEGIN(system_reset)
IVEC=0x100
IAREA=PACA_EXNMI
* Vectors for the FWNMI option. Share common code.
*/
TRAMP_REAL_BEGIN(system_reset_fwnmi)
+ /* XXX: fwnmi guest could run a nested/PR guest, so why no test? */
__IKVM_REAL(system_reset)=0
GEN_INT_ENTRY system_reset, virt=0
GEN_KVM system_reset
+/**
+ * Interrupt 0x200 - Machine Check Interrupt (MCE).
+ * This is a non-maskable interrupt always taken in real-mode. It can be
+ * synchronous or asynchronous, caused by hardware or software, and it may be
+ * taken in a power-saving state.
+ *
+ * Handling:
+ * Similarly to system reset, this uses its own stack and PACA save area,
+ * the difference is re-entrancy is allowed on the machine check stack.
+ *
+ * machine_check_early is run in real mode, and carefully decodes the
+ * machine check and tries to handle it (e.g., flush the SLB if there was an
+ * error detected there), determines if it was recoverable and logs the
+ * event.
+ *
+ * Then, depending on the execution context when the interrupt is taken, there
+ * are 3 main actions:
+ * - Executing in kernel mode. The event is queued with irq_work, which means
+ * it is handled when it is next safe to do so (i.e., the kernel has enabled
+ * interrupts), which could be immediately when the interrupt returns. This
+ * avoids nasty issues like switching to virtual mode when the MMU is in a
+ * bad state, or when executing OPAL code. (SRESET is exposed to such issues,
+ * but it has different priorities). Check to see if the CPU was in power
+ * save, and return via the wake up code if it was.
+ *
+ * - Executing in user mode. machine_check_exception is run like a normal
+ * interrupt handler, which processes the data generated by the early handler.
+ *
+ * - Executing in guest mode. The interrupt is run with its KVM test, and
+ * branches to KVM to deal with. KVM may queue the event for the host
+ * to report later.
+ *
+ * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
+ * or SCRATCH0 is in use, it may cause a crash.
+ *
+ * KVM:
+ * See SRESET.
+ */
INT_DEFINE_BEGIN(machine_check_early)
IVEC=0x200
IAREA=PACA_EXMC
/**
- * 0x300 - Data Storage Interrupt (DSI)
- * This interrupt is generated due to a data access which does not have a valid
- * page table entry with permissions to allow the data access to be performed.
- * DAWR matches also fault here, as do RC updates, and minor misc errors e.g.,
- * copy/paste, AMO, certain invalid CI accesses, etc.
+ * Interrupt 0x300 - Data Storage Interrupt (DSI).
+ * This is a synchronous interrupt generated due to a data access exception,
+ * e.g., a load orstore which does not have a valid page table entry with
+ * permissions. DAWR matches also fault here, as do RC updates, and minor misc
+ * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
+ *
+ * Handling:
+ * - Hash MMU
+ * Go to do_hash_page first to see if the HPT can be filled from an entry in
+ * the Linux page table. Hash faults can hit in kernel mode in a fairly
+ * arbitrary state (e.g., interrupts disabled, locks held) when accessing
+ * "non-bolted" regions, e.g., vmalloc space. However these should always be
+ * backed by Linux page tables.
*
- * This interrupt is delivered to the guest (HV bit unchanged).
+ * If none is found, do a Linux page fault. Linux page faults can happen in
+ * kernel mode due to user copy operations of course.
*
- * Linux HPT responds by first attempting to refill the hash table from the
- * Linux page table, then going to a full page fault if the Linux page table
- * entry was insufficient. RPT goes straight to full page fault.
+ * - Radix MMU
+ * The hardware loads from the Linux page table directly, so a fault goes
+ * immediately to Linux page fault.
*
- * PR KVM ...?
+ * Conditions like DAWR match are handled on the way in to Linux page fault.
*/
INT_DEFINE_BEGIN(data_access)
IVEC=0x300
GEN_KVM data_access
+/**
+ * Interrupt 0x380 - Data Segment Interrupt (DSLB).
+ * This is a synchronous interrupt in response to an MMU fault missing SLB
+ * entry for HPT, or an address outside RPT translation range.
+ *
+ * Handling:
+ * - HPT:
+ * This refills the SLB, or reports an access fault similarly to a bad page
+ * fault. When coming from user-mode, the SLB handler may access any kernel
+ * data, though it may itself take a DSLB. When coming from kernel mode,
+ * recursive faults must be avoided so access is restricted to the kernel
+ * image text/data, kernel stack, and any data allocated below
+ * ppc64_bolted_size (first segment). The kernel handler must avoid stomping
+ * on user-handler data structures.
+ *
+ * A dedicated save area EXSLB is used (XXX: but it actually need not be
+ * these days, we could use EXGEN).
+ */
INT_DEFINE_BEGIN(data_access_slb)
IVEC=0x380
IAREA=PACA_EXSLB
GEN_KVM data_access_slb
+/**
+ * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
+ * This is a synchronous interrupt in response to an MMU fault due to an
+ * instruction fetch.
+ *
+ * Handling:
+ * Similar to DSI, though in response to fetch. The faulting address is found
+ * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
+ */
INT_DEFINE_BEGIN(instruction_access)
IVEC=0x400
IISIDE=1
GEN_KVM instruction_access
+/**
+ * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
+ * This is a synchronous interrupt in response to an MMU fault due to an
+ * instruction fetch.
+ *
+ * Handling:
+ * Similar to DSLB, though in response to fetch. The faulting address is found
+ * in SRR0 (rather than DAR).
+ */
INT_DEFINE_BEGIN(instruction_access_slb)
IVEC=0x480
IAREA=PACA_EXSLB
GEN_KVM instruction_access_slb
+/**
+ * Interrupt 0x500 - External Interrupt.
+ * This is an asynchronous maskable interrupt in response to an "external
+ * exception" from the interrupt controller or hypervisor (e.g., device
+ * interrupt). It is maskable in hardware by clearing MSR[EE], and
+ * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
+ *
+ * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
+ * interrupts are delivered with HSRR registers, guests use SRRs, which
+ * reqiures IHSRR_IF_HVMODE.
+ *
+ * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
+ * external interrupts are delivered as Hypervisor Virtualization Interrupts
+ * rather than External Interrupts.
+ *
+ * Handling:
+ * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
+ * because registers at the time of the interrupt are not so important as it is
+ * asynchronous.
+ *
+ * If soft masked, the masked handler will note the pending interrupt for
+ * replay, and clear MSR[EE] in the interrupted context.
+ */
INT_DEFINE_BEGIN(hardware_interrupt)
IVEC=0x500
IHSRR_IF_HVMODE=1
GEN_KVM hardware_interrupt
+/**
+ * Interrupt 0x600 - Alignment Interrupt
+ * This is a synchronous interrupt in response to data alignment fault.
+ */
INT_DEFINE_BEGIN(alignment)
IVEC=0x600
IDAR=1
GEN_KVM alignment
+/**
+ * Interrupt 0x700 - Program Interrupt (program check).
+ * This is a synchronous interrupt in response to various instruction faults:
+ * traps, privilege errors, TM errors, floating point exceptions.
+ *
+ * Handling:
+ * This interrupt may use the "emergency stack" in some cases when being taken
+ * from kernel context, which complicates handling.
+ */
INT_DEFINE_BEGIN(program_check)
IVEC=0x700
IKVM_REAL=1
GEN_KVM program_check
+/*
+ * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
+ * This is a synchronous interrupt in response to executing an fp instruction
+ * with MSR[FP]=0.
+ *
+ * Handling:
+ * This will load FP registers and enable the FP bit if coming from userspace,
+ * otherwise report a bad kernel use of FP.
+ */
INT_DEFINE_BEGIN(fp_unavailable)
IVEC=0x800
IRECONCILE=0
GEN_KVM fp_unavailable
+/**
+ * Interrupt 0x900 - Decrementer Interrupt.
+ * This is an asynchronous interrupt in response to a decrementer exception
+ * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
+ * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
+ * local_irq_disable()).
+ *
+ * Handling:
+ * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
+ *
+ * If soft masked, the masked handler will note the pending interrupt for
+ * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
+ * in the interrupted context.
+ * If PPC_WATCHDOG is configured, the soft masked handler will actually set
+ * things back up to run soft_nmi_interrupt as a regular interrupt handler
+ * on the emergency stack.
+ */
INT_DEFINE_BEGIN(decrementer)
IVEC=0x900
IMASK=IRQS_DISABLED
GEN_KVM decrementer
+/**
+ * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
+ * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
+ * register.
+ *
+ * Handling:
+ * Linux does not use this outside KVM where it's used to keep a host timer
+ * while the guest is given control of DEC. It should normally be caught by
+ * the KVM test and routed there.
+ */
INT_DEFINE_BEGIN(hdecrementer)
IVEC=0x980
IHSRR=1
GEN_KVM hdecrementer
+/**
+ * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
+ * This is an asynchronous interrupt in response to a msgsndp doorbell.
+ * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
+ * IRQS_DISABLED mask (i.e., local_irq_disable()).
+ *
+ * Handling:
+ * Guests may use this for IPIs between threads in a core if the
+ * hypervisor supports it. NVGPRS are not saved (see 0x500).
+ *
+ * If soft masked, the masked handler will note the pending interrupt for
+ * replay, leaving MSR[EE] enabled in the interrupted context because the
+ * doorbells are edge triggered.
+ */
INT_DEFINE_BEGIN(doorbell_super)
IVEC=0xa00
IMASK=IRQS_DISABLED
EXC_REAL_NONE(0xb00, 0x100)
EXC_VIRT_NONE(0x4b00, 0x100)
-/*
- * system call / hypercall (0xc00, 0x4c00)
- *
- * The system call exception is invoked with "sc 0" and does not alter HV bit.
- *
- * The hypercall is invoked with "sc 1" and sets HV=1.
+/**
+ * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
+ * This is a synchronous interrupt invoked with the "sc" instruction. The
+ * system call is invoked with "sc 0" and does not alter the HV bit, so it
+ * is directed to the currently running OS. The hypercall is invoked with
+ * "sc 1" and it sets HV=1, so it elevates to hypervisor.
*
* In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
* 0x4c00 virtual mode.
*
+ * Handling:
+ * If the KVM test fires then it was due to a hypercall and is accordingly
+ * routed to KVM. Otherwise this executes a normal Linux system call.
+ *
* Call convention:
*
* syscall and hypercalls register conventions are documented in
#endif
+/**
+ * Interrupt 0xd00 - Trace Interrupt.
+ * This is a synchronous interrupt in response to instruction step or
+ * breakpoint faults.
+ */
INT_DEFINE_BEGIN(single_step)
IVEC=0xd00
IKVM_REAL=1
GEN_KVM single_step
+/**
+ * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
+ * This is a synchronous interrupt in response to an MMU fault caused by a
+ * guest data access.
+ *
+ * Handling:
+ * This should always get routed to KVM. In radix MMU mode, this is caused
+ * by a guest nested radix access that can't be performed due to the
+ * partition scope page table. In hash mode, this can be caused by guests
+ * running with translation disabled (virtual real mode) or with VPM enabled.
+ * KVM will update the page table structures or disallow the access.
+ */
INT_DEFINE_BEGIN(h_data_storage)
IVEC=0xe00
IHSRR=1
GEN_KVM h_data_storage
+/**
+ * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
+ * This is a synchronous interrupt in response to an MMU fault caused by a
+ * guest instruction fetch, similar to HDSI.
+ */
INT_DEFINE_BEGIN(h_instr_storage)
IVEC=0xe20
IHSRR=1
GEN_KVM h_instr_storage
+/**
+ * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
+ */
INT_DEFINE_BEGIN(emulation_assist)
IVEC=0xe40
IHSRR=1
GEN_KVM emulation_assist
-/*
- * hmi_exception trampoline is a special case. It jumps to hmi_exception_early
- * first, and then eventaully from there to the trampoline to get into virtual
- * mode.
+/**
+ * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
+ * This is an asynchronous interrupt caused by a Hypervisor Maintenance
+ * Exception. It is always taken in real mode but uses HSRR registers
+ * unlike SRESET and MCE.
+ *
+ * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
+ * with IRQS_DISABLED mask (i.e., local_irq_disable()).
+ *
+ * Handling:
+ * This is a special case, this is handled similarly to machine checks, with an
+ * initial real mode handler that is not soft-masked, which attempts to fix the
+ * problem. Then a regular handler which is soft-maskable and reports the
+ * problem.
+ *
+ * The emergency stack is used for the early real mode handler.
+ *
+ * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
+ * either use soft-masking for the MCE, or use irq_work for the HMI.
+ *
+ * KVM:
+ * Unlike MCE, this calls into KVM without calling the real mode handler
+ * first.
*/
INT_DEFINE_BEGIN(hmi_exception_early)
IVEC=0xe60
GEN_KVM hmi_exception
+/**
+ * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
+ * This is an asynchronous interrupt in response to a msgsnd doorbell.
+ * Similar to the 0xa00 doorbell but for host rather than guest.
+ */
INT_DEFINE_BEGIN(h_doorbell)
IVEC=0xe80
IHSRR=1
GEN_KVM h_doorbell
+/**
+ * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
+ * This is an asynchronous interrupt in response to an "external exception".
+ * Similar to 0x500 but for host only.
+ */
INT_DEFINE_BEGIN(h_virt_irq)
IVEC=0xea0
IHSRR=1
EXC_VIRT_NONE(0x4ee0, 0x20)
+/*
+ * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
+ * This is an asynchronous interrupt in response to a PMU exception.
+ * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
+ * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
+ *
+ * Handling:
+ * This calls into the perf subsystem.
+ *
+ * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
+ * runs under local_irq_disable. However it may be soft-masked in
+ * powerpc-specific code.
+ *
+ * If soft masked, the masked handler will note the pending interrupt for
+ * replay, and clear MSR[EE] in the interrupted context.
+ */
INT_DEFINE_BEGIN(performance_monitor)
IVEC=0xf00
IMASK=IRQS_PMI_DISABLED
GEN_KVM performance_monitor
+/**
+ * Interrupt 0xf20 - Vector Unavailable Interrupt.
+ * This is a synchronous interrupt in response to
+ * executing a vector (or altivec) instruction with MSR[VEC]=0.
+ * Similar to FP unavailable.
+ */
INT_DEFINE_BEGIN(altivec_unavailable)
IVEC=0xf20
IRECONCILE=0
GEN_KVM altivec_unavailable
+/**
+ * Interrupt 0xf40 - VSX Unavailable Interrupt.
+ * This is a synchronous interrupt in response to
+ * executing a VSX instruction with MSR[VSX]=0.
+ * Similar to FP unavailable.
+ */
INT_DEFINE_BEGIN(vsx_unavailable)
IVEC=0xf40
IRECONCILE=0
GEN_KVM vsx_unavailable
+/**
+ * Interrupt 0xf60 - Facility Unavailable Interrupt.
+ * This is a synchronous interrupt in response to
+ * executing an instruction without access to the facility that can be
+ * resolved by the OS (e.g., FSCR, MSR).
+ * Similar to FP unavailable.
+ */
INT_DEFINE_BEGIN(facility_unavailable)
IVEC=0xf60
IKVM_REAL=1
GEN_KVM facility_unavailable
+/**
+ * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
+ * This is a synchronous interrupt in response to
+ * executing an instruction without access to the facility that can only
+ * be resolved in HV mode (e.g., HFSCR).
+ * Similar to FP unavailable.
+ */
INT_DEFINE_BEGIN(h_facility_unavailable)
IVEC=0xf80
IHSRR=1
EXC_REAL_NONE(0x1400, 0x100)
EXC_VIRT_NONE(0x5400, 0x100)
+/**
+ * Interrupt 0x1500 - Soft Patch Interrupt
+ *
+ * Handling:
+ * This is an implementation specific interrupt which can be used for a
+ * range of exceptions.
+ *
+ * This interrupt handler is unique in that it runs the denormal assist
+ * code even for guests (and even in guest context) without going to KVM,
+ * for speed. POWER9 does not raise denorm exceptions, so this special case
+ * could be phased out in future to reduce special cases.
+ */
INT_DEFINE_BEGIN(denorm_exception)
IVEC=0x1500
IHSRR=1
projects / platform / kernel / linux-rpi.git / commitdiff