1
0
Fork 0
alistair23-linux/arch/powerpc/kernel/exceptions-64e.S

1692 lines
45 KiB
ArmAsm
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0-or-later */
/*
* Boot code and exception vectors for Book3E processors
*
* Copyright (C) 2007 Ben. Herrenschmidt (benh@kernel.crashing.org), IBM Corp.
*/
#include <linux/threads.h>
#include <asm/reg.h>
#include <asm/page.h>
#include <asm/ppc_asm.h>
#include <asm/asm-offsets.h>
#include <asm/cputable.h>
#include <asm/setup.h>
#include <asm/thread_info.h>
#include <asm/reg_a2.h>
#include <asm/exception-64e.h>
#include <asm/bug.h>
#include <asm/irqflags.h>
#include <asm/ptrace.h>
#include <asm/ppc-opcode.h>
#include <asm/mmu.h>
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#include <asm/hw_irq.h>
#include <asm/kvm_asm.h>
#include <asm/kvm_booke_hv_asm.h>
#include <asm/feature-fixups.h>
/* XXX This will ultimately add space for a special exception save
* structure used to save things like SRR0/SRR1, SPRGs, MAS, etc...
* when taking special interrupts. For now we don't support that,
* special interrupts from within a non-standard level will probably
* blow you up
*/
#define SPECIAL_EXC_SRR0 0
#define SPECIAL_EXC_SRR1 1
#define SPECIAL_EXC_SPRG_GEN 2
#define SPECIAL_EXC_SPRG_TLB 3
#define SPECIAL_EXC_MAS0 4
#define SPECIAL_EXC_MAS1 5
#define SPECIAL_EXC_MAS2 6
#define SPECIAL_EXC_MAS3 7
#define SPECIAL_EXC_MAS6 8
#define SPECIAL_EXC_MAS7 9
#define SPECIAL_EXC_MAS5 10 /* E.HV only */
#define SPECIAL_EXC_MAS8 11 /* E.HV only */
#define SPECIAL_EXC_IRQHAPPENED 12
#define SPECIAL_EXC_DEAR 13
#define SPECIAL_EXC_ESR 14
#define SPECIAL_EXC_SOFTE 15
#define SPECIAL_EXC_CSRR0 16
#define SPECIAL_EXC_CSRR1 17
/* must be even to keep 16-byte stack alignment */
#define SPECIAL_EXC_END 18
#define SPECIAL_EXC_FRAME_SIZE (INT_FRAME_SIZE + SPECIAL_EXC_END * 8)
#define SPECIAL_EXC_FRAME_OFFS (INT_FRAME_SIZE - 288)
#define SPECIAL_EXC_STORE(reg, name) \
std reg, (SPECIAL_EXC_##name * 8 + SPECIAL_EXC_FRAME_OFFS)(r1)
#define SPECIAL_EXC_LOAD(reg, name) \
ld reg, (SPECIAL_EXC_##name * 8 + SPECIAL_EXC_FRAME_OFFS)(r1)
special_reg_save:
lbz r9,PACAIRQHAPPENED(r13)
RECONCILE_IRQ_STATE(r3,r4)
/*
* We only need (or have stack space) to save this stuff if
* we interrupted the kernel.
*/
ld r3,_MSR(r1)
andi. r3,r3,MSR_PR
bnelr
/*
* Advance to the next TLB exception frame for handler
* types that don't do it automatically.
*/
LOAD_REG_ADDR(r11,extlb_level_exc)
lwz r12,0(r11)
mfspr r10,SPRN_SPRG_TLB_EXFRAME
add r10,r10,r12
mtspr SPRN_SPRG_TLB_EXFRAME,r10
/*
* Save registers needed to allow nesting of certain exceptions
* (such as TLB misses) inside special exception levels
*/
mfspr r10,SPRN_SRR0
SPECIAL_EXC_STORE(r10,SRR0)
mfspr r10,SPRN_SRR1
SPECIAL_EXC_STORE(r10,SRR1)
mfspr r10,SPRN_SPRG_GEN_SCRATCH
SPECIAL_EXC_STORE(r10,SPRG_GEN)
mfspr r10,SPRN_SPRG_TLB_SCRATCH
SPECIAL_EXC_STORE(r10,SPRG_TLB)
mfspr r10,SPRN_MAS0
SPECIAL_EXC_STORE(r10,MAS0)
mfspr r10,SPRN_MAS1
SPECIAL_EXC_STORE(r10,MAS1)
mfspr r10,SPRN_MAS2
SPECIAL_EXC_STORE(r10,MAS2)
mfspr r10,SPRN_MAS3
SPECIAL_EXC_STORE(r10,MAS3)
mfspr r10,SPRN_MAS6
SPECIAL_EXC_STORE(r10,MAS6)
mfspr r10,SPRN_MAS7
SPECIAL_EXC_STORE(r10,MAS7)
BEGIN_FTR_SECTION
mfspr r10,SPRN_MAS5
SPECIAL_EXC_STORE(r10,MAS5)
mfspr r10,SPRN_MAS8
SPECIAL_EXC_STORE(r10,MAS8)
/* MAS5/8 could have inappropriate values if we interrupted KVM code */
li r10,0
mtspr SPRN_MAS5,r10
mtspr SPRN_MAS8,r10
END_FTR_SECTION_IFSET(CPU_FTR_EMB_HV)
SPECIAL_EXC_STORE(r9,IRQHAPPENED)
mfspr r10,SPRN_DEAR
SPECIAL_EXC_STORE(r10,DEAR)
mfspr r10,SPRN_ESR
SPECIAL_EXC_STORE(r10,ESR)
lbz r10,PACAIRQSOFTMASK(r13)
SPECIAL_EXC_STORE(r10,SOFTE)
ld r10,_NIP(r1)
SPECIAL_EXC_STORE(r10,CSRR0)
ld r10,_MSR(r1)
SPECIAL_EXC_STORE(r10,CSRR1)
blr
ret_from_level_except:
ld r3,_MSR(r1)
andi. r3,r3,MSR_PR
beq 1f
b ret_from_except
1:
LOAD_REG_ADDR(r11,extlb_level_exc)
lwz r12,0(r11)
mfspr r10,SPRN_SPRG_TLB_EXFRAME
sub r10,r10,r12
mtspr SPRN_SPRG_TLB_EXFRAME,r10
/*
* It's possible that the special level exception interrupted a
* TLB miss handler, and inserted the same entry that the
* interrupted handler was about to insert. On CPUs without TLB
* write conditional, this can result in a duplicate TLB entry.
* Wipe all non-bolted entries to be safe.
*
* Note that this doesn't protect against any TLB misses
* we may take accessing the stack from here to the end of
* the special level exception. It's not clear how we can
* reasonably protect against that, but only CPUs with
* neither TLB write conditional nor bolted kernel memory
* are affected. Do any such CPUs even exist?
*/
PPC_TLBILX_ALL(0,R0)
REST_NVGPRS(r1)
SPECIAL_EXC_LOAD(r10,SRR0)
mtspr SPRN_SRR0,r10
SPECIAL_EXC_LOAD(r10,SRR1)
mtspr SPRN_SRR1,r10
SPECIAL_EXC_LOAD(r10,SPRG_GEN)
mtspr SPRN_SPRG_GEN_SCRATCH,r10
SPECIAL_EXC_LOAD(r10,SPRG_TLB)
mtspr SPRN_SPRG_TLB_SCRATCH,r10
SPECIAL_EXC_LOAD(r10,MAS0)
mtspr SPRN_MAS0,r10
SPECIAL_EXC_LOAD(r10,MAS1)
mtspr SPRN_MAS1,r10
SPECIAL_EXC_LOAD(r10,MAS2)
mtspr SPRN_MAS2,r10
SPECIAL_EXC_LOAD(r10,MAS3)
mtspr SPRN_MAS3,r10
SPECIAL_EXC_LOAD(r10,MAS6)
mtspr SPRN_MAS6,r10
SPECIAL_EXC_LOAD(r10,MAS7)
mtspr SPRN_MAS7,r10
BEGIN_FTR_SECTION
SPECIAL_EXC_LOAD(r10,MAS5)
mtspr SPRN_MAS5,r10
SPECIAL_EXC_LOAD(r10,MAS8)
mtspr SPRN_MAS8,r10
END_FTR_SECTION_IFSET(CPU_FTR_EMB_HV)
lbz r6,PACAIRQSOFTMASK(r13)
ld r5,SOFTE(r1)
/* Interrupts had better not already be enabled... */
powerpc/64: Change soft_enabled from flag to bitmask "paca->soft_enabled" is used as a flag to mask some of interrupts. Currently supported flags values and their details: soft_enabled MSR[EE] 0 0 Disabled (PMI and HMI not masked) 1 1 Enabled "paca->soft_enabled" is initialized to 1 to make the interripts as enabled. arch_local_irq_disable() will toggle the value when interrupts needs to disbled. At this point, the interrupts are not actually disabled, instead, interrupt vector has code to check for the flag and mask it when it occurs. By "mask it", it update interrupt paca->irq_happened and return. arch_local_irq_restore() is called to re-enable interrupts, which checks and replays interrupts if any occured. Now, as mentioned, current logic doesnot mask "performance monitoring interrupts" and PMIs are implemented as NMI. But this patchset depends on local_irq_* for a successful local_* update. Meaning, mask all possible interrupts during local_* update and replay them after the update. So the idea here is to reserve the "paca->soft_enabled" logic. New values and details: soft_enabled MSR[EE] 1 0 Disabled (PMI and HMI not masked) 0 1 Enabled Reason for the this change is to create foundation for a third mask value "0x2" for "soft_enabled" to add support to mask PMIs. When ->soft_enabled is set to a value "3", PMI interrupts are mask and when set to a value of "1", PMI are not mask. With this patch also extends soft_enabled as interrupt disable mask. Current flags are renamed from IRQ_[EN?DIS}ABLED to IRQS_ENABLED and IRQS_DISABLED. Patch also fixes the ptrace call to force the user to see the softe value to be alway 1. Reason being, even though userspace has no business knowing about softe, it is part of pt_regs. Like-wise in signal context. Signed-off-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-19 20:55:49 -07:00
tweqi r6,IRQS_ENABLED
powerpc/64: Change soft_enabled from flag to bitmask "paca->soft_enabled" is used as a flag to mask some of interrupts. Currently supported flags values and their details: soft_enabled MSR[EE] 0 0 Disabled (PMI and HMI not masked) 1 1 Enabled "paca->soft_enabled" is initialized to 1 to make the interripts as enabled. arch_local_irq_disable() will toggle the value when interrupts needs to disbled. At this point, the interrupts are not actually disabled, instead, interrupt vector has code to check for the flag and mask it when it occurs. By "mask it", it update interrupt paca->irq_happened and return. arch_local_irq_restore() is called to re-enable interrupts, which checks and replays interrupts if any occured. Now, as mentioned, current logic doesnot mask "performance monitoring interrupts" and PMIs are implemented as NMI. But this patchset depends on local_irq_* for a successful local_* update. Meaning, mask all possible interrupts during local_* update and replay them after the update. So the idea here is to reserve the "paca->soft_enabled" logic. New values and details: soft_enabled MSR[EE] 1 0 Disabled (PMI and HMI not masked) 0 1 Enabled Reason for the this change is to create foundation for a third mask value "0x2" for "soft_enabled" to add support to mask PMIs. When ->soft_enabled is set to a value "3", PMI interrupts are mask and when set to a value of "1", PMI are not mask. With this patch also extends soft_enabled as interrupt disable mask. Current flags are renamed from IRQ_[EN?DIS}ABLED to IRQS_ENABLED and IRQS_DISABLED. Patch also fixes the ptrace call to force the user to see the softe value to be alway 1. Reason being, even though userspace has no business knowing about softe, it is part of pt_regs. Like-wise in signal context. Signed-off-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-19 20:55:49 -07:00
andi. r6,r5,IRQS_DISABLED
bne 1f
TRACE_ENABLE_INTS
stb r5,PACAIRQSOFTMASK(r13)
1:
/*
* Restore PACAIRQHAPPENED rather than setting it based on
* the return MSR[EE], since we could have interrupted
* __check_irq_replay() or other inconsistent transitory
* states that must remain that way.
*/
SPECIAL_EXC_LOAD(r10,IRQHAPPENED)
stb r10,PACAIRQHAPPENED(r13)
SPECIAL_EXC_LOAD(r10,DEAR)
mtspr SPRN_DEAR,r10
SPECIAL_EXC_LOAD(r10,ESR)
mtspr SPRN_ESR,r10
stdcx. r0,0,r1 /* to clear the reservation */
REST_4GPRS(2, r1)
REST_4GPRS(6, r1)
ld r10,_CTR(r1)
ld r11,_XER(r1)
mtctr r10
mtxer r11
blr
.macro ret_from_level srr0 srr1 paca_ex scratch
bl ret_from_level_except
ld r10,_LINK(r1)
ld r11,_CCR(r1)
ld r0,GPR13(r1)
mtlr r10
mtcr r11
ld r10,GPR10(r1)
ld r11,GPR11(r1)
ld r12,GPR12(r1)
mtspr \scratch,r0
std r10,\paca_ex+EX_R10(r13);
std r11,\paca_ex+EX_R11(r13);
ld r10,_NIP(r1)
ld r11,_MSR(r1)
ld r0,GPR0(r1)
ld r1,GPR1(r1)
mtspr \srr0,r10
mtspr \srr1,r11
ld r10,\paca_ex+EX_R10(r13)
ld r11,\paca_ex+EX_R11(r13)
mfspr r13,\scratch
.endm
ret_from_crit_except:
ret_from_level SPRN_CSRR0 SPRN_CSRR1 PACA_EXCRIT SPRN_SPRG_CRIT_SCRATCH
rfci
ret_from_mc_except:
ret_from_level SPRN_MCSRR0 SPRN_MCSRR1 PACA_EXMC SPRN_SPRG_MC_SCRATCH
rfmci
/* Exception prolog code for all exceptions */
#define EXCEPTION_PROLOG(n, intnum, type, addition) \
mtspr SPRN_SPRG_##type##_SCRATCH,r13; /* get spare registers */ \
mfspr r13,SPRN_SPRG_PACA; /* get PACA */ \
std r10,PACA_EX##type+EX_R10(r13); \
std r11,PACA_EX##type+EX_R11(r13); \
mfcr r10; /* save CR */ \
mfspr r11,SPRN_##type##_SRR1;/* what are we coming from */ \
DO_KVM intnum,SPRN_##type##_SRR1; /* KVM hook */ \
stw r10,PACA_EX##type+EX_CR(r13); /* save old CR in the PACA */ \
addition; /* additional code for that exc. */ \
std r1,PACA_EX##type+EX_R1(r13); /* save old r1 in the PACA */ \
type##_SET_KSTACK; /* get special stack if necessary */\
andi. r10,r11,MSR_PR; /* save stack pointer */ \
beq 1f; /* branch around if supervisor */ \
ld r1,PACAKSAVE(r13); /* get kernel stack coming from usr */\
1: type##_BTB_FLUSH \
cmpdi cr1,r1,0; /* check if SP makes sense */ \
bge- cr1,exc_##n##_bad_stack;/* bad stack (TODO: out of line) */ \
mfspr r10,SPRN_##type##_SRR0; /* read SRR0 before touching stack */
/* Exception type-specific macros */
#define GEN_SET_KSTACK \
subi r1,r1,INT_FRAME_SIZE; /* alloc frame on kernel stack */
#define SPRN_GEN_SRR0 SPRN_SRR0
#define SPRN_GEN_SRR1 SPRN_SRR1
#define GDBELL_SET_KSTACK GEN_SET_KSTACK
#define SPRN_GDBELL_SRR0 SPRN_GSRR0
#define SPRN_GDBELL_SRR1 SPRN_GSRR1
#define CRIT_SET_KSTACK \
ld r1,PACA_CRIT_STACK(r13); \
subi r1,r1,SPECIAL_EXC_FRAME_SIZE
#define SPRN_CRIT_SRR0 SPRN_CSRR0
#define SPRN_CRIT_SRR1 SPRN_CSRR1
#define DBG_SET_KSTACK \
ld r1,PACA_DBG_STACK(r13); \
subi r1,r1,SPECIAL_EXC_FRAME_SIZE
#define SPRN_DBG_SRR0 SPRN_DSRR0
#define SPRN_DBG_SRR1 SPRN_DSRR1
#define MC_SET_KSTACK \
ld r1,PACA_MC_STACK(r13); \
subi r1,r1,SPECIAL_EXC_FRAME_SIZE
#define SPRN_MC_SRR0 SPRN_MCSRR0
#define SPRN_MC_SRR1 SPRN_MCSRR1
#ifdef CONFIG_PPC_FSL_BOOK3E
#define GEN_BTB_FLUSH \
START_BTB_FLUSH_SECTION \
beq 1f; \
BTB_FLUSH(r10) \
1: \
END_BTB_FLUSH_SECTION
#define CRIT_BTB_FLUSH \
START_BTB_FLUSH_SECTION \
BTB_FLUSH(r10) \
END_BTB_FLUSH_SECTION
#define DBG_BTB_FLUSH CRIT_BTB_FLUSH
#define MC_BTB_FLUSH CRIT_BTB_FLUSH
#define GDBELL_BTB_FLUSH GEN_BTB_FLUSH
#else
#define GEN_BTB_FLUSH
#define CRIT_BTB_FLUSH
#define DBG_BTB_FLUSH
#define MC_BTB_FLUSH
#define GDBELL_BTB_FLUSH
#endif
#define NORMAL_EXCEPTION_PROLOG(n, intnum, addition) \
EXCEPTION_PROLOG(n, intnum, GEN, addition##_GEN(n))
#define CRIT_EXCEPTION_PROLOG(n, intnum, addition) \
EXCEPTION_PROLOG(n, intnum, CRIT, addition##_CRIT(n))
#define DBG_EXCEPTION_PROLOG(n, intnum, addition) \
EXCEPTION_PROLOG(n, intnum, DBG, addition##_DBG(n))
#define MC_EXCEPTION_PROLOG(n, intnum, addition) \
EXCEPTION_PROLOG(n, intnum, MC, addition##_MC(n))
#define GDBELL_EXCEPTION_PROLOG(n, intnum, addition) \
EXCEPTION_PROLOG(n, intnum, GDBELL, addition##_GDBELL(n))
/* Variants of the "addition" argument for the prolog
*/
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_NONE_GEN(n)
#define PROLOG_ADDITION_NONE_GDBELL(n)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_NONE_CRIT(n)
#define PROLOG_ADDITION_NONE_DBG(n)
#define PROLOG_ADDITION_NONE_MC(n)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_MASKABLE_GEN(n) \
lbz r10,PACAIRQSOFTMASK(r13); /* are irqs soft-masked? */ \
powerpc/64: Change soft_enabled from flag to bitmask "paca->soft_enabled" is used as a flag to mask some of interrupts. Currently supported flags values and their details: soft_enabled MSR[EE] 0 0 Disabled (PMI and HMI not masked) 1 1 Enabled "paca->soft_enabled" is initialized to 1 to make the interripts as enabled. arch_local_irq_disable() will toggle the value when interrupts needs to disbled. At this point, the interrupts are not actually disabled, instead, interrupt vector has code to check for the flag and mask it when it occurs. By "mask it", it update interrupt paca->irq_happened and return. arch_local_irq_restore() is called to re-enable interrupts, which checks and replays interrupts if any occured. Now, as mentioned, current logic doesnot mask "performance monitoring interrupts" and PMIs are implemented as NMI. But this patchset depends on local_irq_* for a successful local_* update. Meaning, mask all possible interrupts during local_* update and replay them after the update. So the idea here is to reserve the "paca->soft_enabled" logic. New values and details: soft_enabled MSR[EE] 1 0 Disabled (PMI and HMI not masked) 0 1 Enabled Reason for the this change is to create foundation for a third mask value "0x2" for "soft_enabled" to add support to mask PMIs. When ->soft_enabled is set to a value "3", PMI interrupts are mask and when set to a value of "1", PMI are not mask. With this patch also extends soft_enabled as interrupt disable mask. Current flags are renamed from IRQ_[EN?DIS}ABLED to IRQS_ENABLED and IRQS_DISABLED. Patch also fixes the ptrace call to force the user to see the softe value to be alway 1. Reason being, even though userspace has no business knowing about softe, it is part of pt_regs. Like-wise in signal context. Signed-off-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-19 20:55:49 -07:00
andi. r10,r10,IRQS_DISABLED; /* yes -> go out of line */ \
bne masked_interrupt_book3e_##n
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_2REGS_GEN(n) \
std r14,PACA_EXGEN+EX_R14(r13); \
std r15,PACA_EXGEN+EX_R15(r13)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_1REG_GEN(n) \
std r14,PACA_EXGEN+EX_R14(r13);
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_2REGS_CRIT(n) \
std r14,PACA_EXCRIT+EX_R14(r13); \
std r15,PACA_EXCRIT+EX_R15(r13)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_2REGS_DBG(n) \
std r14,PACA_EXDBG+EX_R14(r13); \
std r15,PACA_EXDBG+EX_R15(r13)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
#define PROLOG_ADDITION_2REGS_MC(n) \
std r14,PACA_EXMC+EX_R14(r13); \
std r15,PACA_EXMC+EX_R15(r13)
/* Core exception code for all exceptions except TLB misses. */
#define EXCEPTION_COMMON_LVL(n, scratch, excf) \
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
exc_##n##_common: \
std r0,GPR0(r1); /* save r0 in stackframe */ \
std r2,GPR2(r1); /* save r2 in stackframe */ \
SAVE_4GPRS(3, r1); /* save r3 - r6 in stackframe */ \
SAVE_2GPRS(7, r1); /* save r7, r8 in stackframe */ \
std r9,GPR9(r1); /* save r9 in stackframe */ \
std r10,_NIP(r1); /* save SRR0 to stackframe */ \
std r11,_MSR(r1); /* save SRR1 to stackframe */ \
beq 2f; /* if from kernel mode */ \
ACCOUNT_CPU_USER_ENTRY(r13,r10,r11);/* accounting (uses cr0+eq) */ \
2: ld r3,excf+EX_R10(r13); /* get back r10 */ \
ld r4,excf+EX_R11(r13); /* get back r11 */ \
mfspr r5,scratch; /* get back r13 */ \
std r12,GPR12(r1); /* save r12 in stackframe */ \
ld r2,PACATOC(r13); /* get kernel TOC into r2 */ \
mflr r6; /* save LR in stackframe */ \
mfctr r7; /* save CTR in stackframe */ \
mfspr r8,SPRN_XER; /* save XER in stackframe */ \
ld r9,excf+EX_R1(r13); /* load orig r1 back from PACA */ \
lwz r10,excf+EX_CR(r13); /* load orig CR back from PACA */ \
lbz r11,PACAIRQSOFTMASK(r13); /* get current IRQ softe */ \
ld r12,exception_marker@toc(r2); \
li r0,0; \
std r3,GPR10(r1); /* save r10 to stackframe */ \
std r4,GPR11(r1); /* save r11 to stackframe */ \
std r5,GPR13(r1); /* save it to stackframe */ \
std r6,_LINK(r1); \
std r7,_CTR(r1); \
std r8,_XER(r1); \
li r3,(n)+1; /* indicate partial regs in trap */ \
std r9,0(r1); /* store stack frame back link */ \
std r10,_CCR(r1); /* store orig CR in stackframe */ \
std r9,GPR1(r1); /* store stack frame back link */ \
std r11,SOFTE(r1); /* and save it to stackframe */ \
std r12,STACK_FRAME_OVERHEAD-16(r1); /* mark the frame */ \
std r3,_TRAP(r1); /* set trap number */ \
std r0,RESULT(r1); /* clear regs->result */
#define EXCEPTION_COMMON(n) \
EXCEPTION_COMMON_LVL(n, SPRN_SPRG_GEN_SCRATCH, PACA_EXGEN)
#define EXCEPTION_COMMON_CRIT(n) \
EXCEPTION_COMMON_LVL(n, SPRN_SPRG_CRIT_SCRATCH, PACA_EXCRIT)
#define EXCEPTION_COMMON_MC(n) \
EXCEPTION_COMMON_LVL(n, SPRN_SPRG_MC_SCRATCH, PACA_EXMC)
#define EXCEPTION_COMMON_DBG(n) \
EXCEPTION_COMMON_LVL(n, SPRN_SPRG_DBG_SCRATCH, PACA_EXDBG)
/*
* This is meant for exceptions that don't immediately hard-enable. We
* set a bit in paca->irq_happened to ensure that a subsequent call to
* arch_local_irq_restore() will properly hard-enable and avoid the
* fast-path, and then reconcile irq state.
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
*/
#define INTS_DISABLE RECONCILE_IRQ_STATE(r3,r4)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/*
* This is called by exceptions that don't use INTS_DISABLE (that did not
* touch irq indicators in the PACA). This will restore MSR:EE to it's
* previous value
*
* XXX In the long run, we may want to open-code it in order to separate the
* load from the wrtee, thus limiting the latency caused by the dependency
* but at this point, I'll favor code clarity until we have a near to final
* implementation
*/
#define INTS_RESTORE_HARD \
ld r11,_MSR(r1); \
wrtee r11;
/* XXX FIXME: Restore r14/r15 when necessary */
#define BAD_STACK_TRAMPOLINE(n) \
exc_##n##_bad_stack: \
li r1,(n); /* get exception number */ \
sth r1,PACA_TRAP_SAVE(r13); /* store trap */ \
b bad_stack_book3e; /* bad stack error */
/* WARNING: If you change the layout of this stub, make sure you check
* the debug exception handler which handles single stepping
* into exceptions from userspace, and the MM code in
* arch/powerpc/mm/tlb_nohash.c which patches the branch here
* and would need to be updated if that branch is moved
*/
#define EXCEPTION_STUB(loc, label) \
. = interrupt_base_book3e + loc; \
nop; /* To make debug interrupts happy */ \
b exc_##label##_book3e;
#define ACK_NONE(r)
#define ACK_DEC(r) \
lis r,TSR_DIS@h; \
mtspr SPRN_TSR,r
#define ACK_FIT(r) \
lis r,TSR_FIS@h; \
mtspr SPRN_TSR,r
/* Used by asynchronous interrupt that may happen in the idle loop.
*
* This check if the thread was in the idle loop, and if yes, returns
* to the caller rather than the PC. This is to avoid a race if
* interrupts happen before the wait instruction.
*/
#define CHECK_NAPPING() \
ld r11, PACA_THREAD_INFO(r13); \
ld r10,TI_LOCAL_FLAGS(r11); \
andi. r9,r10,_TLF_NAPPING; \
beq+ 1f; \
ld r8,_LINK(r1); \
rlwinm r7,r10,0,~_TLF_NAPPING; \
std r8,_NIP(r1); \
std r7,TI_LOCAL_FLAGS(r11); \
1:
#define MASKABLE_EXCEPTION(trapnum, intnum, label, hdlr, ack) \
START_EXCEPTION(label); \
NORMAL_EXCEPTION_PROLOG(trapnum, intnum, PROLOG_ADDITION_MASKABLE)\
EXCEPTION_COMMON(trapnum) \
INTS_DISABLE; \
ack(r8); \
CHECK_NAPPING(); \
addi r3,r1,STACK_FRAME_OVERHEAD; \
bl hdlr; \
b ret_from_except_lite;
/* This value is used to mark exception frames on the stack. */
.section ".toc","aw"
exception_marker:
.tc ID_EXC_MARKER[TC],STACK_FRAME_REGS_MARKER
/*
* And here we have the exception vectors !
*/
.text
.balign 0x1000
.globl interrupt_base_book3e
interrupt_base_book3e: /* fake trap */
EXCEPTION_STUB(0x000, machine_check)
EXCEPTION_STUB(0x020, critical_input) /* 0x0100 */
EXCEPTION_STUB(0x040, debug_crit) /* 0x0d00 */
EXCEPTION_STUB(0x060, data_storage) /* 0x0300 */
EXCEPTION_STUB(0x080, instruction_storage) /* 0x0400 */
EXCEPTION_STUB(0x0a0, external_input) /* 0x0500 */
EXCEPTION_STUB(0x0c0, alignment) /* 0x0600 */
EXCEPTION_STUB(0x0e0, program) /* 0x0700 */
EXCEPTION_STUB(0x100, fp_unavailable) /* 0x0800 */
EXCEPTION_STUB(0x120, system_call) /* 0x0c00 */
EXCEPTION_STUB(0x140, ap_unavailable) /* 0x0f20 */
EXCEPTION_STUB(0x160, decrementer) /* 0x0900 */
EXCEPTION_STUB(0x180, fixed_interval) /* 0x0980 */
EXCEPTION_STUB(0x1a0, watchdog) /* 0x09f0 */
EXCEPTION_STUB(0x1c0, data_tlb_miss)
EXCEPTION_STUB(0x1e0, instruction_tlb_miss)
EXCEPTION_STUB(0x200, altivec_unavailable)
EXCEPTION_STUB(0x220, altivec_assist)
EXCEPTION_STUB(0x260, perfmon)
EXCEPTION_STUB(0x280, doorbell)
EXCEPTION_STUB(0x2a0, doorbell_crit)
EXCEPTION_STUB(0x2c0, guest_doorbell)
EXCEPTION_STUB(0x2e0, guest_doorbell_crit)
EXCEPTION_STUB(0x300, hypercall)
EXCEPTION_STUB(0x320, ehpriv)
EXCEPTION_STUB(0x340, lrat_error)
.globl __end_interrupts
__end_interrupts:
/* Critical Input Interrupt */
START_EXCEPTION(critical_input);
CRIT_EXCEPTION_PROLOG(0x100, BOOKE_INTERRUPT_CRITICAL,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON_CRIT(0x100)
bl save_nvgprs
bl special_reg_save
CHECK_NAPPING();
addi r3,r1,STACK_FRAME_OVERHEAD
bl unknown_exception
b ret_from_crit_except
/* Machine Check Interrupt */
START_EXCEPTION(machine_check);
MC_EXCEPTION_PROLOG(0x000, BOOKE_INTERRUPT_MACHINE_CHECK,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON_MC(0x000)
bl save_nvgprs
bl special_reg_save
CHECK_NAPPING();
addi r3,r1,STACK_FRAME_OVERHEAD
bl machine_check_exception
b ret_from_mc_except
/* Data Storage Interrupt */
START_EXCEPTION(data_storage)
NORMAL_EXCEPTION_PROLOG(0x300, BOOKE_INTERRUPT_DATA_STORAGE,
PROLOG_ADDITION_2REGS)
mfspr r14,SPRN_DEAR
mfspr r15,SPRN_ESR
EXCEPTION_COMMON(0x300)
INTS_DISABLE
b storage_fault_common
/* Instruction Storage Interrupt */
START_EXCEPTION(instruction_storage);
NORMAL_EXCEPTION_PROLOG(0x400, BOOKE_INTERRUPT_INST_STORAGE,
PROLOG_ADDITION_2REGS)
li r15,0
mr r14,r10
EXCEPTION_COMMON(0x400)
INTS_DISABLE
b storage_fault_common
/* External Input Interrupt */
MASKABLE_EXCEPTION(0x500, BOOKE_INTERRUPT_EXTERNAL,
external_input, do_IRQ, ACK_NONE)
/* Alignment */
START_EXCEPTION(alignment);
NORMAL_EXCEPTION_PROLOG(0x600, BOOKE_INTERRUPT_ALIGNMENT,
PROLOG_ADDITION_2REGS)
mfspr r14,SPRN_DEAR
mfspr r15,SPRN_ESR
EXCEPTION_COMMON(0x600)
b alignment_more /* no room, go out of line */
/* Program Interrupt */
START_EXCEPTION(program);
NORMAL_EXCEPTION_PROLOG(0x700, BOOKE_INTERRUPT_PROGRAM,
PROLOG_ADDITION_1REG)
mfspr r14,SPRN_ESR
EXCEPTION_COMMON(0x700)
INTS_DISABLE
std r14,_DSISR(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
ld r14,PACA_EXGEN+EX_R14(r13)
bl save_nvgprs
bl program_check_exception
b ret_from_except
/* Floating Point Unavailable Interrupt */
START_EXCEPTION(fp_unavailable);
NORMAL_EXCEPTION_PROLOG(0x800, BOOKE_INTERRUPT_FP_UNAVAIL,
PROLOG_ADDITION_NONE)
/* we can probably do a shorter exception entry for that one... */
EXCEPTION_COMMON(0x800)
ld r12,_MSR(r1)
andi. r0,r12,MSR_PR;
beq- 1f
bl load_up_fpu
b fast_exception_return
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
1: INTS_DISABLE
bl save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
bl kernel_fp_unavailable_exception
b ret_from_except
/* Altivec Unavailable Interrupt */
START_EXCEPTION(altivec_unavailable);
NORMAL_EXCEPTION_PROLOG(0x200, BOOKE_INTERRUPT_ALTIVEC_UNAVAIL,
PROLOG_ADDITION_NONE)
/* we can probably do a shorter exception entry for that one... */
EXCEPTION_COMMON(0x200)
#ifdef CONFIG_ALTIVEC
BEGIN_FTR_SECTION
ld r12,_MSR(r1)
andi. r0,r12,MSR_PR;
beq- 1f
bl load_up_altivec
b fast_exception_return
1:
END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
#endif
INTS_DISABLE
bl save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
bl altivec_unavailable_exception
b ret_from_except
/* AltiVec Assist */
START_EXCEPTION(altivec_assist);
NORMAL_EXCEPTION_PROLOG(0x220,
BOOKE_INTERRUPT_ALTIVEC_ASSIST,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0x220)
INTS_DISABLE
bl save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
#ifdef CONFIG_ALTIVEC
BEGIN_FTR_SECTION
bl altivec_assist_exception
END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
#else
bl unknown_exception
#endif
b ret_from_except
/* Decrementer Interrupt */
MASKABLE_EXCEPTION(0x900, BOOKE_INTERRUPT_DECREMENTER,
decrementer, timer_interrupt, ACK_DEC)
/* Fixed Interval Timer Interrupt */
MASKABLE_EXCEPTION(0x980, BOOKE_INTERRUPT_FIT,
fixed_interval, unknown_exception, ACK_FIT)
/* Watchdog Timer Interrupt */
START_EXCEPTION(watchdog);
CRIT_EXCEPTION_PROLOG(0x9f0, BOOKE_INTERRUPT_WATCHDOG,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON_CRIT(0x9f0)
bl save_nvgprs
bl special_reg_save
CHECK_NAPPING();
addi r3,r1,STACK_FRAME_OVERHEAD
#ifdef CONFIG_BOOKE_WDT
bl WatchdogException
#else
bl unknown_exception
#endif
b ret_from_crit_except
/* System Call Interrupt */
START_EXCEPTION(system_call)
mr r9,r13 /* keep a copy of userland r13 */
mfspr r11,SPRN_SRR0 /* get return address */
mfspr r12,SPRN_SRR1 /* get previous MSR */
mfspr r13,SPRN_SPRG_PACA /* get our PACA */
b system_call_common
/* Auxiliary Processor Unavailable Interrupt */
START_EXCEPTION(ap_unavailable);
NORMAL_EXCEPTION_PROLOG(0xf20, BOOKE_INTERRUPT_AP_UNAVAIL,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0xf20)
INTS_DISABLE
bl save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
bl unknown_exception
b ret_from_except
/* Debug exception as a critical interrupt*/
START_EXCEPTION(debug_crit);
CRIT_EXCEPTION_PROLOG(0xd00, BOOKE_INTERRUPT_DEBUG,
PROLOG_ADDITION_2REGS)
/*
* If there is a single step or branch-taken exception in an
* exception entry sequence, it was probably meant to apply to
* the code where the exception occurred (since exception entry
* doesn't turn off DE automatically). We simulate the effect
* of turning off DE on entry to an exception handler by turning
* off DE in the CSRR1 value and clearing the debug status.
*/
mfspr r14,SPRN_DBSR /* check single-step/branch taken */
andis. r15,r14,(DBSR_IC|DBSR_BT)@h
beq+ 1f
powerpc/64e: Fix hang when debugging programs with relocated kernel Debug interrupts can be taken during interrupt entry, since interrupt entry does not automatically turn them off. The kernel will check whether the faulting instruction is between [interrupt_base_book3e, __end_interrupts], and if so clear MSR[DE] and return. However, when the kernel is built with CONFIG_RELOCATABLE, it can't use LOAD_REG_IMMEDIATE(r14,interrupt_base_book3e) and LOAD_REG_IMMEDIATE(r15,__end_interrupts), as they ignore relocation. Thus, if the kernel is actually running at a different address than it was built at, the address comparison will fail, and the exception entry code will hang at kernel_dbg_exc. r2(toc) is also not usable here, as r2 still holds data from the interrupted context, so LOAD_REG_ADDR() doesn't work either. So we use the *name@got* to get the EV of two labels directly. Test programs test.c shows as follows: int main(int argc, char *argv[]) { if (access("/proc/sys/kernel/perf_event_paranoid", F_OK) == -1) printf("Kernel doesn't have perf_event support\n"); } Steps to reproduce the bug, for example: 1) ./gdb ./test 2) (gdb) b access 3) (gdb) r 4) (gdb) s Signed-off-by: Liu Hailong <liu.hailong6@zte.com.cn> Signed-off-by: Jiang Xuexin <jiang.xuexin@zte.com.cn> Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn> Reviewed-by: Liu Song <liu.song11@zte.com.cn> Reviewed-by: Huang Jian <huang.jian@zte.com.cn> [scottwood: cleaned up commit message, and specified bad behavior as a hang rather than an oops to correspond to mainline kernel behavior] Fixes: 1cb6e0649248 ("powerpc/book3e: support CONFIG_RELOCATABLE") Cc: <stable@vger.kernel.org> # 4.4.x- Signed-off-by: Scott Wood <oss@buserror.net>
2017-02-06 19:35:52 -07:00
#ifdef CONFIG_RELOCATABLE
ld r15,PACATOC(r13)
ld r14,interrupt_base_book3e@got(r15)
ld r15,__end_interrupts@got(r15)
cmpld cr0,r10,r14
cmpld cr1,r10,r15
#else
LOAD_REG_IMMEDIATE_SYM(r14, r15, interrupt_base_book3e)
cmpld cr0, r10, r14
LOAD_REG_IMMEDIATE_SYM(r14, r15, __end_interrupts)
cmpld cr1, r10, r14
#endif
blt+ cr0,1f
bge+ cr1,1f
/* here it looks like we got an inappropriate debug exception. */
lis r14,(DBSR_IC|DBSR_BT)@h /* clear the event */
rlwinm r11,r11,0,~MSR_DE /* clear DE in the CSRR1 value */
mtspr SPRN_DBSR,r14
mtspr SPRN_CSRR1,r11
lwz r10,PACA_EXCRIT+EX_CR(r13) /* restore registers */
ld r1,PACA_EXCRIT+EX_R1(r13)
ld r14,PACA_EXCRIT+EX_R14(r13)
ld r15,PACA_EXCRIT+EX_R15(r13)
mtcr r10
ld r10,PACA_EXCRIT+EX_R10(r13) /* restore registers */
ld r11,PACA_EXCRIT+EX_R11(r13)
powerpc/booke64: Use SPRG7 for VDSO Previously SPRG3 was marked for use by both VDSO and critical interrupts (though critical interrupts were not fully implemented). In commit 8b64a9dfb091f1eca8b7e58da82f1e7d1d5fe0ad ("powerpc/booke64: Use SPRG0/3 scratch for bolted TLB miss & crit int"), Mihai Caraman made an attempt to resolve this conflict by restoring the VDSO value early in the critical interrupt, but this has some issues: - It's incompatible with EXCEPTION_COMMON which restores r13 from the by-then-overwritten scratch (this cost me some debugging time). - It forces critical exceptions to be a special case handled differently from even machine check and debug level exceptions. - It didn't occur to me that it was possible to make this work at all (by doing a final "ld r13, PACA_EXCRIT+EX_R13(r13)") until after I made (most of) this patch. :-) It might be worth investigating using a load rather than SPRG on return from all exceptions (except TLB misses where the scratch never leaves the SPRG) -- it could save a few cycles. Until then, let's stick with SPRG for all exceptions. Since we cannot use SPRG4-7 for scratch without corrupting the state of a KVM guest, move VDSO to SPRG7 on book3e. Since neither SPRG4-7 nor critical interrupts exist on book3s, SPRG3 is still used for VDSO there. Signed-off-by: Scott Wood <scottwood@freescale.com> Cc: Mihai Caraman <mihai.caraman@freescale.com> Cc: Anton Blanchard <anton@samba.org> Cc: Paul Mackerras <paulus@samba.org> Cc: kvm-ppc@vger.kernel.org
2014-03-10 16:29:38 -06:00
mfspr r13,SPRN_SPRG_CRIT_SCRATCH
rfci
/* Normal debug exception */
/* XXX We only handle coming from userspace for now since we can't
* quite save properly an interrupted kernel state yet
*/
1: andi. r14,r11,MSR_PR; /* check for userspace again */
beq kernel_dbg_exc; /* if from kernel mode */
/* Now we mash up things to make it look like we are coming on a
* normal exception
*/
mfspr r14,SPRN_DBSR
EXCEPTION_COMMON_CRIT(0xd00)
std r14,_DSISR(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
mr r4,r14
ld r14,PACA_EXCRIT+EX_R14(r13)
ld r15,PACA_EXCRIT+EX_R15(r13)
bl save_nvgprs
bl DebugException
b ret_from_except
kernel_dbg_exc:
b . /* NYI */
/* Debug exception as a debug interrupt*/
START_EXCEPTION(debug_debug);
DBG_EXCEPTION_PROLOG(0xd00, BOOKE_INTERRUPT_DEBUG,
PROLOG_ADDITION_2REGS)
/*
* If there is a single step or branch-taken exception in an
* exception entry sequence, it was probably meant to apply to
* the code where the exception occurred (since exception entry
* doesn't turn off DE automatically). We simulate the effect
* of turning off DE on entry to an exception handler by turning
* off DE in the DSRR1 value and clearing the debug status.
*/
mfspr r14,SPRN_DBSR /* check single-step/branch taken */
andis. r15,r14,(DBSR_IC|DBSR_BT)@h
beq+ 1f
powerpc/64e: Fix hang when debugging programs with relocated kernel Debug interrupts can be taken during interrupt entry, since interrupt entry does not automatically turn them off. The kernel will check whether the faulting instruction is between [interrupt_base_book3e, __end_interrupts], and if so clear MSR[DE] and return. However, when the kernel is built with CONFIG_RELOCATABLE, it can't use LOAD_REG_IMMEDIATE(r14,interrupt_base_book3e) and LOAD_REG_IMMEDIATE(r15,__end_interrupts), as they ignore relocation. Thus, if the kernel is actually running at a different address than it was built at, the address comparison will fail, and the exception entry code will hang at kernel_dbg_exc. r2(toc) is also not usable here, as r2 still holds data from the interrupted context, so LOAD_REG_ADDR() doesn't work either. So we use the *name@got* to get the EV of two labels directly. Test programs test.c shows as follows: int main(int argc, char *argv[]) { if (access("/proc/sys/kernel/perf_event_paranoid", F_OK) == -1) printf("Kernel doesn't have perf_event support\n"); } Steps to reproduce the bug, for example: 1) ./gdb ./test 2) (gdb) b access 3) (gdb) r 4) (gdb) s Signed-off-by: Liu Hailong <liu.hailong6@zte.com.cn> Signed-off-by: Jiang Xuexin <jiang.xuexin@zte.com.cn> Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn> Reviewed-by: Liu Song <liu.song11@zte.com.cn> Reviewed-by: Huang Jian <huang.jian@zte.com.cn> [scottwood: cleaned up commit message, and specified bad behavior as a hang rather than an oops to correspond to mainline kernel behavior] Fixes: 1cb6e0649248 ("powerpc/book3e: support CONFIG_RELOCATABLE") Cc: <stable@vger.kernel.org> # 4.4.x- Signed-off-by: Scott Wood <oss@buserror.net>
2017-02-06 19:35:52 -07:00
#ifdef CONFIG_RELOCATABLE
ld r15,PACATOC(r13)
ld r14,interrupt_base_book3e@got(r15)
ld r15,__end_interrupts@got(r15)
cmpld cr0,r10,r14
cmpld cr1,r10,r15
#else
LOAD_REG_IMMEDIATE_SYM(r14, r15, interrupt_base_book3e)
cmpld cr0, r10, r14
LOAD_REG_IMMEDIATE_SYM(r14, r15,__end_interrupts)
cmpld cr1, r10, r14
#endif
blt+ cr0,1f
bge+ cr1,1f
/* here it looks like we got an inappropriate debug exception. */
lis r14,(DBSR_IC|DBSR_BT)@h /* clear the event */
rlwinm r11,r11,0,~MSR_DE /* clear DE in the DSRR1 value */
mtspr SPRN_DBSR,r14
mtspr SPRN_DSRR1,r11
lwz r10,PACA_EXDBG+EX_CR(r13) /* restore registers */
ld r1,PACA_EXDBG+EX_R1(r13)
ld r14,PACA_EXDBG+EX_R14(r13)
ld r15,PACA_EXDBG+EX_R15(r13)
mtcr r10
ld r10,PACA_EXDBG+EX_R10(r13) /* restore registers */
ld r11,PACA_EXDBG+EX_R11(r13)
mfspr r13,SPRN_SPRG_DBG_SCRATCH
rfdi
/* Normal debug exception */
/* XXX We only handle coming from userspace for now since we can't
* quite save properly an interrupted kernel state yet
*/
1: andi. r14,r11,MSR_PR; /* check for userspace again */
beq kernel_dbg_exc; /* if from kernel mode */
/* Now we mash up things to make it look like we are coming on a
* normal exception
*/
mfspr r14,SPRN_DBSR
EXCEPTION_COMMON_DBG(0xd08)
INTS_DISABLE
std r14,_DSISR(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
mr r4,r14
ld r14,PACA_EXDBG+EX_R14(r13)
ld r15,PACA_EXDBG+EX_R15(r13)
bl save_nvgprs
bl DebugException
b ret_from_except
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
START_EXCEPTION(perfmon);
NORMAL_EXCEPTION_PROLOG(0x260, BOOKE_INTERRUPT_PERFORMANCE_MONITOR,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0x260)
INTS_DISABLE
CHECK_NAPPING()
addi r3,r1,STACK_FRAME_OVERHEAD
bl performance_monitor_exception
b ret_from_except_lite
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/* Doorbell interrupt */
MASKABLE_EXCEPTION(0x280, BOOKE_INTERRUPT_DOORBELL,
doorbell, doorbell_exception, ACK_NONE)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/* Doorbell critical Interrupt */
START_EXCEPTION(doorbell_crit);
CRIT_EXCEPTION_PROLOG(0x2a0, BOOKE_INTERRUPT_DOORBELL_CRITICAL,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON_CRIT(0x2a0)
bl save_nvgprs
bl special_reg_save
CHECK_NAPPING();
addi r3,r1,STACK_FRAME_OVERHEAD
bl unknown_exception
b ret_from_crit_except
/*
* Guest doorbell interrupt
* This general exception use GSRRx save/restore registers
*/
START_EXCEPTION(guest_doorbell);
GDBELL_EXCEPTION_PROLOG(0x2c0, BOOKE_INTERRUPT_GUEST_DBELL,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0x2c0)
addi r3,r1,STACK_FRAME_OVERHEAD
bl save_nvgprs
INTS_RESTORE_HARD
bl unknown_exception
b ret_from_except
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/* Guest Doorbell critical Interrupt */
START_EXCEPTION(guest_doorbell_crit);
CRIT_EXCEPTION_PROLOG(0x2e0, BOOKE_INTERRUPT_GUEST_DBELL_CRIT,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON_CRIT(0x2e0)
bl save_nvgprs
bl special_reg_save
CHECK_NAPPING();
addi r3,r1,STACK_FRAME_OVERHEAD
bl unknown_exception
b ret_from_crit_except
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/* Hypervisor call */
START_EXCEPTION(hypercall);
NORMAL_EXCEPTION_PROLOG(0x310, BOOKE_INTERRUPT_HV_SYSCALL,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0x310)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
addi r3,r1,STACK_FRAME_OVERHEAD
bl save_nvgprs
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
INTS_RESTORE_HARD
bl unknown_exception
b ret_from_except
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/* Embedded Hypervisor priviledged */
START_EXCEPTION(ehpriv);
NORMAL_EXCEPTION_PROLOG(0x320, BOOKE_INTERRUPT_HV_PRIV,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0x320)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
addi r3,r1,STACK_FRAME_OVERHEAD
bl save_nvgprs
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
INTS_RESTORE_HARD
bl unknown_exception
b ret_from_except
/* LRAT Error interrupt */
START_EXCEPTION(lrat_error);
NORMAL_EXCEPTION_PROLOG(0x340, BOOKE_INTERRUPT_LRAT_ERROR,
PROLOG_ADDITION_NONE)
EXCEPTION_COMMON(0x340)
addi r3,r1,STACK_FRAME_OVERHEAD
bl save_nvgprs
INTS_RESTORE_HARD
bl unknown_exception
b ret_from_except
/*
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
* An interrupt came in while soft-disabled; We mark paca->irq_happened
* accordingly and if the interrupt is level sensitive, we hard disable
powerpc/64s: Fix may_hard_irq_enable() for PMI soft masking The soft IRQ masking code has to hard-disable interrupts in cases where the exception is not cleared by the masked handler. External interrupts used this approach for soft masking. Now recently PMU interrupts do the same thing. The soft IRQ masking code additionally allowed for interrupt handlers to hard-enable interrupts after soft-disabling them. The idea is to allow PMU interrupts through to profile interrupt handlers. So when interrupts are being replayed when there is a pending interrupt that requires hard-disabling, there is a test to prevent those handlers from hard-enabling them if there is a pending external interrupt. may_hard_irq_enable() handles this. After f442d00480 ("powerpc/64s: Add support to mask perf interrupts and replay them"), may_hard_irq_enable() could prematurely enable MSR[EE] when a PMU exception exists, which would result in the interrupt firing again while masked, and MSR[EE] being disabled again. I haven't seen that this could cause a serious problem, but it's more consistent to handle these soft-masked interrupts in the same way. So introduce a define for all types of interrupts that require MSR[EE] masking in their soft-disable handlers, and use that in may_hard_irq_enable(). Fixes: f442d004806e ("powerpc/64s: Add support to mask perf interrupts and replay them") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Reviewed-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-02-03 00:17:50 -07:00
* hard disable (full_mask) corresponds to PACA_IRQ_MUST_HARD_MASK, so
* keep these in synch.
*/
.macro masked_interrupt_book3e paca_irq full_mask
lbz r10,PACAIRQHAPPENED(r13)
.if \full_mask == 1
ori r10,r10,\paca_irq | PACA_IRQ_HARD_DIS
.else
ori r10,r10,\paca_irq
.endif
stb r10,PACAIRQHAPPENED(r13)
.if \full_mask == 1
rldicl r10,r11,48,1 /* clear MSR_EE */
rotldi r11,r10,16
mtspr SPRN_SRR1,r11
.endif
lwz r11,PACA_EXGEN+EX_CR(r13)
mtcr r11
ld r10,PACA_EXGEN+EX_R10(r13)
ld r11,PACA_EXGEN+EX_R11(r13)
mfspr r13,SPRN_SPRG_GEN_SCRATCH
rfi
b .
.endm
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
masked_interrupt_book3e_0x500:
// XXX When adding support for EPR, use PACA_IRQ_EE_EDGE
masked_interrupt_book3e PACA_IRQ_EE 1
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
masked_interrupt_book3e_0x900:
ACK_DEC(r10);
masked_interrupt_book3e PACA_IRQ_DEC 0
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
masked_interrupt_book3e_0x980:
ACK_FIT(r10);
masked_interrupt_book3e PACA_IRQ_DEC 0
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
masked_interrupt_book3e_0x280:
masked_interrupt_book3e_0x2c0:
masked_interrupt_book3e PACA_IRQ_DBELL 0
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
/*
* Called from arch_local_irq_enable when an interrupt needs
* to be resent. r3 contains either 0x500,0x900,0x260 or 0x280
* to indicate the kind of interrupt. MSR:EE is already off.
* We generate a stackframe like if a real interrupt had happened.
*
* Note: While MSR:EE is off, we need to make sure that _MSR
* in the generated frame has EE set to 1 or the exception
* handler will not properly re-enable them.
*/
_GLOBAL(__replay_interrupt)
/* We are going to jump to the exception common code which
* will retrieve various register values from the PACA which
* we don't give a damn about.
*/
mflr r10
mfmsr r11
mfcr r4
mtspr SPRN_SPRG_GEN_SCRATCH,r13;
std r1,PACA_EXGEN+EX_R1(r13);
stw r4,PACA_EXGEN+EX_CR(r13);
ori r11,r11,MSR_EE
subi r1,r1,INT_FRAME_SIZE;
cmpwi cr0,r3,0x500
beq exc_0x500_common
cmpwi cr0,r3,0x900
beq exc_0x900_common
cmpwi cr0,r3,0x280
beq exc_0x280_common
blr
/*
* This is called from 0x300 and 0x400 handlers after the prologs with
* r14 and r15 containing the fault address and error code, with the
* original values stashed away in the PACA
*/
storage_fault_common:
std r14,_DAR(r1)
std r15,_DSISR(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
mr r4,r14
mr r5,r15
ld r14,PACA_EXGEN+EX_R14(r13)
ld r15,PACA_EXGEN+EX_R15(r13)
bl do_page_fault
cmpdi r3,0
bne- 1f
b ret_from_except_lite
1: bl save_nvgprs
mr r5,r3
addi r3,r1,STACK_FRAME_OVERHEAD
ld r4,_DAR(r1)
bl bad_page_fault
b ret_from_except
/*
* Alignment exception doesn't fit entirely in the 0x100 bytes so it
* continues here.
*/
alignment_more:
std r14,_DAR(r1)
std r15,_DSISR(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
ld r14,PACA_EXGEN+EX_R14(r13)
ld r15,PACA_EXGEN+EX_R15(r13)
bl save_nvgprs
INTS_RESTORE_HARD
bl alignment_exception
b ret_from_except
/*
* We branch here from entry_64.S for the last stage of the exception
* return code path. MSR:EE is expected to be off at that point
*/
_GLOBAL(exception_return_book3e)
b 1f
/* This is the return from load_up_fpu fast path which could do with
* less GPR restores in fact, but for now we have a single return path
*/
.globl fast_exception_return
fast_exception_return:
wrteei 0
1: mr r0,r13
ld r10,_MSR(r1)
REST_4GPRS(2, r1)
andi. r6,r10,MSR_PR
REST_2GPRS(6, r1)
beq 1f
ACCOUNT_CPU_USER_EXIT(r13, r10, r11)
ld r0,GPR13(r1)
1: stdcx. r0,0,r1 /* to clear the reservation */
ld r8,_CCR(r1)
ld r9,_LINK(r1)
ld r10,_CTR(r1)
ld r11,_XER(r1)
mtcr r8
mtlr r9
mtctr r10
mtxer r11
REST_2GPRS(8, r1)
ld r10,GPR10(r1)
ld r11,GPR11(r1)
ld r12,GPR12(r1)
mtspr SPRN_SPRG_GEN_SCRATCH,r0
std r10,PACA_EXGEN+EX_R10(r13);
std r11,PACA_EXGEN+EX_R11(r13);
ld r10,_NIP(r1)
ld r11,_MSR(r1)
ld r0,GPR0(r1)
ld r1,GPR1(r1)
mtspr SPRN_SRR0,r10
mtspr SPRN_SRR1,r11
ld r10,PACA_EXGEN+EX_R10(r13)
ld r11,PACA_EXGEN+EX_R11(r13)
mfspr r13,SPRN_SPRG_GEN_SCRATCH
rfi
/*
* Trampolines used when spotting a bad kernel stack pointer in
* the exception entry code.
*
* TODO: move some bits like SRR0 read to trampoline, pass PACA
* index around, etc... to handle crit & mcheck
*/
BAD_STACK_TRAMPOLINE(0x000)
BAD_STACK_TRAMPOLINE(0x100)
BAD_STACK_TRAMPOLINE(0x200)
BAD_STACK_TRAMPOLINE(0x220)
BAD_STACK_TRAMPOLINE(0x260)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
BAD_STACK_TRAMPOLINE(0x280)
BAD_STACK_TRAMPOLINE(0x2a0)
BAD_STACK_TRAMPOLINE(0x2c0)
BAD_STACK_TRAMPOLINE(0x2e0)
BAD_STACK_TRAMPOLINE(0x300)
BAD_STACK_TRAMPOLINE(0x310)
BAD_STACK_TRAMPOLINE(0x320)
BAD_STACK_TRAMPOLINE(0x340)
BAD_STACK_TRAMPOLINE(0x400)
BAD_STACK_TRAMPOLINE(0x500)
BAD_STACK_TRAMPOLINE(0x600)
BAD_STACK_TRAMPOLINE(0x700)
BAD_STACK_TRAMPOLINE(0x800)
BAD_STACK_TRAMPOLINE(0x900)
BAD_STACK_TRAMPOLINE(0x980)
BAD_STACK_TRAMPOLINE(0x9f0)
BAD_STACK_TRAMPOLINE(0xa00)
BAD_STACK_TRAMPOLINE(0xb00)
BAD_STACK_TRAMPOLINE(0xc00)
BAD_STACK_TRAMPOLINE(0xd00)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 00:27:59 -07:00
BAD_STACK_TRAMPOLINE(0xd08)
BAD_STACK_TRAMPOLINE(0xe00)
BAD_STACK_TRAMPOLINE(0xf00)
BAD_STACK_TRAMPOLINE(0xf20)
.globl bad_stack_book3e
bad_stack_book3e:
/* XXX: Needs to make SPRN_SPRG_GEN depend on exception type */
mfspr r10,SPRN_SRR0; /* read SRR0 before touching stack */
ld r1,PACAEMERGSP(r13)
subi r1,r1,64+INT_FRAME_SIZE
std r10,_NIP(r1)
std r11,_MSR(r1)
ld r10,PACA_EXGEN+EX_R1(r13) /* FIXME for crit & mcheck */
lwz r11,PACA_EXGEN+EX_CR(r13) /* FIXME for crit & mcheck */
std r10,GPR1(r1)
std r11,_CCR(r1)
mfspr r10,SPRN_DEAR
mfspr r11,SPRN_ESR
std r10,_DAR(r1)
std r11,_DSISR(r1)
std r0,GPR0(r1); /* save r0 in stackframe */ \
std r2,GPR2(r1); /* save r2 in stackframe */ \
SAVE_4GPRS(3, r1); /* save r3 - r6 in stackframe */ \
SAVE_2GPRS(7, r1); /* save r7, r8 in stackframe */ \
std r9,GPR9(r1); /* save r9 in stackframe */ \
ld r3,PACA_EXGEN+EX_R10(r13);/* get back r10 */ \
ld r4,PACA_EXGEN+EX_R11(r13);/* get back r11 */ \
mfspr r5,SPRN_SPRG_GEN_SCRATCH;/* get back r13 XXX can be wrong */ \
std r3,GPR10(r1); /* save r10 to stackframe */ \
std r4,GPR11(r1); /* save r11 to stackframe */ \
std r12,GPR12(r1); /* save r12 in stackframe */ \
std r5,GPR13(r1); /* save it to stackframe */ \
mflr r10
mfctr r11
mfxer r12
std r10,_LINK(r1)
std r11,_CTR(r1)
std r12,_XER(r1)
SAVE_10GPRS(14,r1)
SAVE_8GPRS(24,r1)
lhz r12,PACA_TRAP_SAVE(r13)
std r12,_TRAP(r1)
addi r11,r1,INT_FRAME_SIZE
std r11,0(r1)
li r12,0
std r12,0(r11)
ld r2,PACATOC(r13)
1: addi r3,r1,STACK_FRAME_OVERHEAD
bl kernel_bad_stack
b 1b
/*
* Setup the initial TLB for a core. This current implementation
* assume that whatever we are running off will not conflict with
* the new mapping at PAGE_OFFSET.
*/
_GLOBAL(initial_tlb_book3e)
/* Look for the first TLB with IPROT set */
mfspr r4,SPRN_TLB0CFG
andi. r3,r4,TLBnCFG_IPROT
lis r3,MAS0_TLBSEL(0)@h
bne found_iprot
mfspr r4,SPRN_TLB1CFG
andi. r3,r4,TLBnCFG_IPROT
lis r3,MAS0_TLBSEL(1)@h
bne found_iprot
mfspr r4,SPRN_TLB2CFG
andi. r3,r4,TLBnCFG_IPROT
lis r3,MAS0_TLBSEL(2)@h
bne found_iprot
lis r3,MAS0_TLBSEL(3)@h
mfspr r4,SPRN_TLB3CFG
/* fall through */
found_iprot:
andi. r5,r4,TLBnCFG_HES
bne have_hes
mflr r8 /* save LR */
/* 1. Find the index of the entry we're executing in
*
* r3 = MAS0_TLBSEL (for the iprot array)
* r4 = SPRN_TLBnCFG
*/
bl invstr /* Find our address */
invstr: mflr r6 /* Make it accessible */
mfmsr r7
rlwinm r5,r7,27,31,31 /* extract MSR[IS] */
mfspr r7,SPRN_PID
slwi r7,r7,16
or r7,r7,r5
mtspr SPRN_MAS6,r7
tlbsx 0,r6 /* search MSR[IS], SPID=PID */
mfspr r3,SPRN_MAS0
rlwinm r5,r3,16,20,31 /* Extract MAS0(Entry) */
mfspr r7,SPRN_MAS1 /* Insure IPROT set */
oris r7,r7,MAS1_IPROT@h
mtspr SPRN_MAS1,r7
tlbwe
/* 2. Invalidate all entries except the entry we're executing in
*
* r3 = MAS0 w/TLBSEL & ESEL for the entry we are running in
* r4 = SPRN_TLBnCFG
* r5 = ESEL of entry we are running in
*/
andi. r4,r4,TLBnCFG_N_ENTRY /* Extract # entries */
li r6,0 /* Set Entry counter to 0 */
1: mr r7,r3 /* Set MAS0(TLBSEL) */
rlwimi r7,r6,16,4,15 /* Setup MAS0 = TLBSEL | ESEL(r6) */
mtspr SPRN_MAS0,r7
tlbre
mfspr r7,SPRN_MAS1
rlwinm r7,r7,0,2,31 /* Clear MAS1 Valid and IPROT */
cmpw r5,r6
beq skpinv /* Dont update the current execution TLB */
mtspr SPRN_MAS1,r7
tlbwe
isync
skpinv: addi r6,r6,1 /* Increment */
cmpw r6,r4 /* Are we done? */
bne 1b /* If not, repeat */
/* Invalidate all TLBs */
PPC_TLBILX_ALL(0,R0)
sync
isync
/* 3. Setup a temp mapping and jump to it
*
* r3 = MAS0 w/TLBSEL & ESEL for the entry we are running in
* r5 = ESEL of entry we are running in
*/
andi. r7,r5,0x1 /* Find an entry not used and is non-zero */
addi r7,r7,0x1
mr r4,r3 /* Set MAS0(TLBSEL) = 1 */
mtspr SPRN_MAS0,r4
tlbre
rlwimi r4,r7,16,4,15 /* Setup MAS0 = TLBSEL | ESEL(r7) */
mtspr SPRN_MAS0,r4
mfspr r7,SPRN_MAS1
xori r6,r7,MAS1_TS /* Setup TMP mapping in the other Address space */
mtspr SPRN_MAS1,r6
tlbwe
mfmsr r6
xori r6,r6,MSR_IS
mtspr SPRN_SRR1,r6
bl 1f /* Find our address */
1: mflr r6
addi r6,r6,(2f - 1b)
mtspr SPRN_SRR0,r6
rfi
2:
/* 4. Clear out PIDs & Search info
*
* r3 = MAS0 w/TLBSEL & ESEL for the entry we started in
* r4 = MAS0 w/TLBSEL & ESEL for the temp mapping
* r5 = MAS3
*/
li r6,0
mtspr SPRN_MAS6,r6
mtspr SPRN_PID,r6
/* 5. Invalidate mapping we started in
*
* r3 = MAS0 w/TLBSEL & ESEL for the entry we started in
* r4 = MAS0 w/TLBSEL & ESEL for the temp mapping
* r5 = MAS3
*/
mtspr SPRN_MAS0,r3
tlbre
mfspr r6,SPRN_MAS1
rlwinm r6,r6,0,2,31 /* clear IPROT and VALID */
mtspr SPRN_MAS1,r6
tlbwe
sync
isync
/*
* The mapping only needs to be cache-coherent on SMP, except on
* Freescale e500mc derivatives where it's also needed for coherent DMA.
*/
#if defined(CONFIG_SMP) || defined(CONFIG_PPC_E500MC)
#define M_IF_NEEDED MAS2_M
#else
#define M_IF_NEEDED 0
#endif
/* 6. Setup KERNELBASE mapping in TLB[0]
*
* r3 = MAS0 w/TLBSEL & ESEL for the entry we started in
* r4 = MAS0 w/TLBSEL & ESEL for the temp mapping
* r5 = MAS3
*/
rlwinm r3,r3,0,16,3 /* clear ESEL */
mtspr SPRN_MAS0,r3
lis r6,(MAS1_VALID|MAS1_IPROT)@h
ori r6,r6,(MAS1_TSIZE(BOOK3E_PAGESZ_1GB))@l
mtspr SPRN_MAS1,r6
LOAD_REG_IMMEDIATE(r6, PAGE_OFFSET | M_IF_NEEDED)
mtspr SPRN_MAS2,r6
rlwinm r5,r5,0,0,25
ori r5,r5,MAS3_SR | MAS3_SW | MAS3_SX
mtspr SPRN_MAS3,r5
li r5,-1
rlwinm r5,r5,0,0,25
tlbwe
/* 7. Jump to KERNELBASE mapping
*
* r4 = MAS0 w/TLBSEL & ESEL for the temp mapping
*/
/* Now we branch the new virtual address mapped by this entry */
bl 1f /* Find our address */
1: mflr r6
addi r6,r6,(2f - 1b)
tovirt(r6,r6)
lis r7,MSR_KERNEL@h
ori r7,r7,MSR_KERNEL@l
mtspr SPRN_SRR0,r6
mtspr SPRN_SRR1,r7
rfi /* start execution out of TLB1[0] entry */
2:
/* 8. Clear out the temp mapping
*
* r4 = MAS0 w/TLBSEL & ESEL for the entry we are running in
*/
mtspr SPRN_MAS0,r4
tlbre
mfspr r5,SPRN_MAS1
rlwinm r5,r5,0,2,31 /* clear IPROT and VALID */
mtspr SPRN_MAS1,r5
tlbwe
sync
isync
/* We translate LR and return */
tovirt(r8,r8)
mtlr r8
blr
have_hes:
/* Setup MAS 0,1,2,3 and 7 for tlbwe of a 1G entry that maps the
* kernel linear mapping. We also set MAS8 once for all here though
* that will have to be made dependent on whether we are running under
* a hypervisor I suppose.
*/
/* BEWARE, MAGIC
* This code is called as an ordinary function on the boot CPU. But to
* avoid duplication, this code is also used in SCOM bringup of
* secondary CPUs. We read the code between the initial_tlb_code_start
* and initial_tlb_code_end labels one instruction at a time and RAM it
* into the new core via SCOM. That doesn't process branches, so there
* must be none between those two labels. It also means if this code
* ever takes any parameters, the SCOM code must also be updated to
* provide them.
*/
.globl a2_tlbinit_code_start
a2_tlbinit_code_start:
ori r11,r3,MAS0_WQ_ALLWAYS
oris r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 erratum 376 */
mtspr SPRN_MAS0,r11
lis r3,(MAS1_VALID | MAS1_IPROT)@h
ori r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT
mtspr SPRN_MAS1,r3
LOAD_REG_IMMEDIATE(r3, PAGE_OFFSET | MAS2_M)
mtspr SPRN_MAS2,r3
li r3,MAS3_SR | MAS3_SW | MAS3_SX
mtspr SPRN_MAS7_MAS3,r3
li r3,0
mtspr SPRN_MAS8,r3
/* Write the TLB entry */
tlbwe
.globl a2_tlbinit_after_linear_map
a2_tlbinit_after_linear_map:
/* Now we branch the new virtual address mapped by this entry */
LOAD_REG_IMMEDIATE_SYM(r3, r5, 1f)
mtctr r3
bctr
1: /* We are now running at PAGE_OFFSET, clean the TLB of everything
* else (including IPROTed things left by firmware)
* r4 = TLBnCFG
* r3 = current address (more or less)
*/
li r5,0
mtspr SPRN_MAS6,r5
tlbsx 0,r3
rlwinm r9,r4,0,TLBnCFG_N_ENTRY
rlwinm r10,r4,8,0xff
addi r10,r10,-1 /* Get inner loop mask */
li r3,1
mfspr r5,SPRN_MAS1
rlwinm r5,r5,0,(~(MAS1_VALID|MAS1_IPROT))
mfspr r6,SPRN_MAS2
rldicr r6,r6,0,51 /* Extract EPN */
mfspr r7,SPRN_MAS0
rlwinm r7,r7,0,0xffff0fff /* Clear HES and WQ */
rlwinm r8,r7,16,0xfff /* Extract ESEL */
2: add r4,r3,r8
and r4,r4,r10
rlwimi r7,r4,16,MAS0_ESEL_MASK
mtspr SPRN_MAS0,r7
mtspr SPRN_MAS1,r5
mtspr SPRN_MAS2,r6
tlbwe
addi r3,r3,1
and. r4,r3,r10
bne 3f
addis r6,r6,(1<<30)@h
3:
cmpw r3,r9
blt 2b
.globl a2_tlbinit_after_iprot_flush
a2_tlbinit_after_iprot_flush:
PPC_TLBILX(0,0,R0)
sync
isync
.globl a2_tlbinit_code_end
a2_tlbinit_code_end:
/* We translate LR and return */
mflr r3
tovirt(r3,r3)
mtlr r3
blr
/*
* Main entry (boot CPU, thread 0)
*
* We enter here from head_64.S, possibly after the prom_init trampoline
* with r3 and r4 already saved to r31 and 30 respectively and in 64 bits
* mode. Anything else is as it was left by the bootloader
*
* Initial requirements of this port:
*
* - Kernel loaded at 0 physical
* - A good lump of memory mapped 0:0 by UTLB entry 0
* - MSR:IS & MSR:DS set to 0
*
* Note that some of the above requirements will be relaxed in the future
* as the kernel becomes smarter at dealing with different initial conditions
* but for now you have to be careful
*/
_GLOBAL(start_initialization_book3e)
mflr r28
/* First, we need to setup some initial TLBs to map the kernel
* text, data and bss at PAGE_OFFSET. We don't have a real mode
* and always use AS 0, so we just set it up to match our link
* address and never use 0 based addresses.
*/
bl initial_tlb_book3e
/* Init global core bits */
bl init_core_book3e
/* Init per-thread bits */
bl init_thread_book3e
/* Return to common init code */
tovirt(r28,r28)
mtlr r28
blr
/*
* Secondary core/processor entry
*
* This is entered for thread 0 of a secondary core, all other threads
* are expected to be stopped. It's similar to start_initialization_book3e
* except that it's generally entered from the holding loop in head_64.S
* after CPUs have been gathered by Open Firmware.
*
* We assume we are in 32 bits mode running with whatever TLB entry was
* set for us by the firmware or POR engine.
*/
_GLOBAL(book3e_secondary_core_init_tlb_set)
li r4,1
b generic_secondary_smp_init
_GLOBAL(book3e_secondary_core_init)
mflr r28
/* Do we need to setup initial TLB entry ? */
cmplwi r4,0
bne 2f
/* Setup TLB for this core */
bl initial_tlb_book3e
/* We can return from the above running at a different
* address, so recalculate r2 (TOC)
*/
bl relative_toc
/* Init global core bits */
2: bl init_core_book3e
/* Init per-thread bits */
3: bl init_thread_book3e
/* Return to common init code at proper virtual address.
*
* Due to various previous assumptions, we know we entered this
* function at either the final PAGE_OFFSET mapping or using a
* 1:1 mapping at 0, so we don't bother doing a complicated check
* here, we just ensure the return address has the right top bits.
*
* Note that if we ever want to be smarter about where we can be
* started from, we have to be careful that by the time we reach
* the code below we may already be running at a different location
* than the one we were called from since initial_tlb_book3e can
* have moved us already.
*/
cmpdi cr0,r28,0
blt 1f
lis r3,PAGE_OFFSET@highest
sldi r3,r3,32
or r28,r28,r3
1: mtlr r28
blr
_GLOBAL(book3e_secondary_thread_init)
mflr r28
b 3b
.globl init_core_book3e
init_core_book3e:
/* Establish the interrupt vector base */
tovirt(r2,r2)
LOAD_REG_ADDR(r3, interrupt_base_book3e)
mtspr SPRN_IVPR,r3
sync
blr
init_thread_book3e:
lis r3,(SPRN_EPCR_ICM | SPRN_EPCR_GICM)@h
mtspr SPRN_EPCR,r3
/* Make sure interrupts are off */
wrteei 0
/* disable all timers and clear out status */
li r3,0
mtspr SPRN_TCR,r3
mfspr r3,SPRN_TSR
mtspr SPRN_TSR,r3
blr
_GLOBAL(__setup_base_ivors)
SET_IVOR(0, 0x020) /* Critical Input */
SET_IVOR(1, 0x000) /* Machine Check */
SET_IVOR(2, 0x060) /* Data Storage */
SET_IVOR(3, 0x080) /* Instruction Storage */
SET_IVOR(4, 0x0a0) /* External Input */
SET_IVOR(5, 0x0c0) /* Alignment */
SET_IVOR(6, 0x0e0) /* Program */
SET_IVOR(7, 0x100) /* FP Unavailable */
SET_IVOR(8, 0x120) /* System Call */
SET_IVOR(9, 0x140) /* Auxiliary Processor Unavailable */
SET_IVOR(10, 0x160) /* Decrementer */
SET_IVOR(11, 0x180) /* Fixed Interval Timer */
SET_IVOR(12, 0x1a0) /* Watchdog Timer */
SET_IVOR(13, 0x1c0) /* Data TLB Error */
SET_IVOR(14, 0x1e0) /* Instruction TLB Error */
SET_IVOR(15, 0x040) /* Debug */
sync
blr
_GLOBAL(setup_altivec_ivors)
SET_IVOR(32, 0x200) /* AltiVec Unavailable */
SET_IVOR(33, 0x220) /* AltiVec Assist */
blr
_GLOBAL(setup_perfmon_ivor)
SET_IVOR(35, 0x260) /* Performance Monitor */
blr
_GLOBAL(setup_doorbell_ivors)
SET_IVOR(36, 0x280) /* Processor Doorbell */
SET_IVOR(37, 0x2a0) /* Processor Doorbell Crit */
blr
_GLOBAL(setup_ehv_ivors)
SET_IVOR(40, 0x300) /* Embedded Hypervisor System Call */
SET_IVOR(41, 0x320) /* Embedded Hypervisor Privilege */
SET_IVOR(38, 0x2c0) /* Guest Processor Doorbell */
SET_IVOR(39, 0x2e0) /* Guest Processor Doorbell Crit/MC */
blr
_GLOBAL(setup_lrat_ivor)
SET_IVOR(42, 0x340) /* LRAT Error */
blr