diff options
Diffstat (limited to 'arch/xtensa/kernel')
27 files changed, 3058 insertions, 1332 deletions
diff --git a/arch/xtensa/kernel/.gitignore b/arch/xtensa/kernel/.gitignore new file mode 100644 index 00000000000..c5f676c3c22 --- /dev/null +++ b/arch/xtensa/kernel/.gitignore @@ -0,0 +1 @@ +vmlinux.lds diff --git a/arch/xtensa/kernel/Makefile b/arch/xtensa/kernel/Makefile index 59fc3fe1557..18d962a8c0c 100644 --- a/arch/xtensa/kernel/Makefile +++ b/arch/xtensa/kernel/Makefile @@ -4,13 +4,17 @@ extra-y := head.o vmlinux.lds -obj-y := align.o entry.o irq.o coprocessor.o process.o ptrace.o \ - setup.o signal.o syscall.o time.o traps.o vectors.o platform.o \ - pci-dma.o io.o +obj-y := align.o coprocessor.o entry.o irq.o pci-dma.o platform.o process.o \ + ptrace.o setup.o signal.o stacktrace.o syscall.o time.o traps.o \ + vectors.o obj-$(CONFIG_KGDB) += xtensa-stub.o obj-$(CONFIG_PCI) += pci.o obj-$(CONFIG_MODULES) += xtensa_ksyms.o module.o +obj-$(CONFIG_FUNCTION_TRACER) += mcount.o +obj-$(CONFIG_SMP) += smp.o mxhead.o + +AFLAGS_head.o += -mtext-section-literals # In the Xtensa architecture, assembly generates literals which must always # precede the L32R instruction with a relative offset less than 256 kB. @@ -23,12 +27,13 @@ obj-$(CONFIG_MODULES) += xtensa_ksyms.o module.o # # Replicate rules in scripts/Makefile.build -sed-y = -e 's/\*(\(\.[a-z]*it\|\.ref\|\)\.text)/*(\1.literal \1.text)/g' \ +sed-y = -e 's/\*(\(\.[a-z]*it\|\.ref\|\)\.text)/*(\1.literal \1.text)/g' \ + -e 's/\.text\.unlikely/.literal.unlikely .text.unlikely/g' \ -e 's/\*(\(\.text\.[a-z]*\))/*(\1.literal \1)/g' quiet_cmd__cpp_lds_S = LDS $@ - cmd__cpp_lds_S = $(CPP) $(cpp_flags) -P -C -Uxtensa -D__ASSEMBLY__ $< \ - | sed $(sed-y) >$@ +cmd__cpp_lds_S = $(CPP) $(cpp_flags) -P -C -Uxtensa -D__ASSEMBLY__ $< \ + | sed $(sed-y) >$@ $(obj)/vmlinux.lds: $(src)/vmlinux.lds.S FORCE $(call if_changed_dep,_cpp_lds_S) diff --git a/arch/xtensa/kernel/align.S b/arch/xtensa/kernel/align.S index 33d6e9d2e83..d4cef6039a5 100644 --- a/arch/xtensa/kernel/align.S +++ b/arch/xtensa/kernel/align.S @@ -146,9 +146,9 @@ * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -170,15 +170,14 @@ ENTRY(fast_unaligned) s32i a7, a2, PT_AREG7 s32i a8, a2, PT_AREG8 - rsr a0, DEPC - xsr a3, EXCSAVE_1 + rsr a0, depc s32i a0, a2, PT_AREG2 s32i a3, a2, PT_AREG3 /* Keep value of SAR in a0 */ - rsr a0, SAR - rsr a8, EXCVADDR # load unaligned memory address + rsr a0, sar + rsr a8, excvaddr # load unaligned memory address /* Now, identify one of the following load/store instructions. * @@ -197,7 +196,7 @@ ENTRY(fast_unaligned) /* Extract the instruction that caused the unaligned access. */ - rsr a7, EPC_1 # load exception address + rsr a7, epc1 # load exception address movi a3, ~3 and a3, a3, a7 # mask lower bits @@ -275,16 +274,16 @@ ENTRY(fast_unaligned) 1: #if XCHAL_HAVE_LOOPS - rsr a5, LEND # check if we reached LEND + rsr a5, lend # check if we reached LEND bne a7, a5, 1f - rsr a5, LCOUNT # and LCOUNT != 0 + rsr a5, lcount # and LCOUNT != 0 beqz a5, 1f addi a5, a5, -1 # decrement LCOUNT and set - rsr a7, LBEG # set PC to LBEGIN - wsr a5, LCOUNT + rsr a7, lbeg # set PC to LBEGIN + wsr a5, lcount #endif -1: wsr a7, EPC_1 # skip load instruction +1: wsr a7, epc1 # skip load instruction extui a4, a4, INSN_T, 4 # extract target register movi a5, .Lload_table addx8 a4, a4, a5 @@ -355,16 +354,16 @@ ENTRY(fast_unaligned) 1: #if XCHAL_HAVE_LOOPS - rsr a4, LEND # check if we reached LEND + rsr a4, lend # check if we reached LEND bne a7, a4, 1f - rsr a4, LCOUNT # and LCOUNT != 0 + rsr a4, lcount # and LCOUNT != 0 beqz a4, 1f addi a4, a4, -1 # decrement LCOUNT and set - rsr a7, LBEG # set PC to LBEGIN - wsr a4, LCOUNT + rsr a7, lbeg # set PC to LBEGIN + wsr a4, lcount #endif -1: wsr a7, EPC_1 # skip store instruction +1: wsr a7, epc1 # skip store instruction movi a4, ~3 and a4, a4, a8 # align memory address @@ -406,7 +405,7 @@ ENTRY(fast_unaligned) .Lexit: movi a4, 0 - rsr a3, EXCSAVE_1 + rsr a3, excsave1 s32i a4, a3, EXC_TABLE_FIXUP /* Restore working register */ @@ -420,7 +419,7 @@ ENTRY(fast_unaligned) /* restore SAR and return */ - wsr a0, SAR + wsr a0, sar l32i a0, a2, PT_AREG0 l32i a2, a2, PT_AREG2 rfe @@ -438,11 +437,11 @@ ENTRY(fast_unaligned) l32i a6, a2, PT_AREG6 l32i a5, a2, PT_AREG5 l32i a4, a2, PT_AREG4 - wsr a0, SAR + wsr a0, sar mov a1, a2 - rsr a0, PS - bbsi.l a2, PS_UM_BIT, 1f # jump if user mode + rsr a0, ps + bbsi.l a2, PS_UM_BIT, 1f # jump if user mode movi a0, _kernel_exception jx a0 @@ -450,6 +449,6 @@ ENTRY(fast_unaligned) 1: movi a0, _user_exception jx a0 +ENDPROC(fast_unaligned) #endif /* XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION */ - diff --git a/arch/xtensa/kernel/asm-offsets.c b/arch/xtensa/kernel/asm-offsets.c index 7dc3f915718..1915c7c889b 100644 --- a/arch/xtensa/kernel/asm-offsets.c +++ b/arch/xtensa/kernel/asm-offsets.c @@ -41,6 +41,8 @@ int main(void) DEFINE(PT_SAR, offsetof (struct pt_regs, sar)); DEFINE(PT_ICOUNTLEVEL, offsetof (struct pt_regs, icountlevel)); DEFINE(PT_SYSCALL, offsetof (struct pt_regs, syscall)); + DEFINE(PT_SCOMPARE1, offsetof(struct pt_regs, scompare1)); + DEFINE(PT_THREADPTR, offsetof(struct pt_regs, threadptr)); DEFINE(PT_AREG, offsetof (struct pt_regs, areg[0])); DEFINE(PT_AREG0, offsetof (struct pt_regs, areg[0])); DEFINE(PT_AREG1, offsetof (struct pt_regs, areg[1])); @@ -91,7 +93,8 @@ int main(void) #endif DEFINE(THREAD_XTREGS_USER, offsetof (struct thread_info, xtregs_user)); DEFINE(XTREGS_USER_SIZE, sizeof(xtregs_user_t)); - DEFINE(THREAD_CURRENT_DS, offsetof (struct task_struct, thread.current_ds)); + DEFINE(THREAD_CURRENT_DS, offsetof (struct task_struct, \ + thread.current_ds)); /* struct mm_struct */ DEFINE(MM_USERS, offsetof(struct mm_struct, mm_users)); @@ -108,4 +111,3 @@ int main(void) return 0; } - diff --git a/arch/xtensa/kernel/coprocessor.S b/arch/xtensa/kernel/coprocessor.S index 2bc1e145c0a..a482df5df2b 100644 --- a/arch/xtensa/kernel/coprocessor.S +++ b/arch/xtensa/kernel/coprocessor.S @@ -32,9 +32,9 @@ * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -43,10 +43,13 @@ /* IO protection is currently unsupported. */ ENTRY(fast_io_protect) - wsr a0, EXCSAVE_1 + + wsr a0, excsave1 movi a0, unrecoverable_exception callx0 a0 +ENDPROC(fast_io_protect) + #if XTENSA_HAVE_COPROCESSORS /* @@ -139,6 +142,7 @@ ENTRY(fast_io_protect) */ ENTRY(coprocessor_save) + entry a1, 32 s32i a0, a1, 0 movi a0, .Lsave_cp_regs_jump_table @@ -150,7 +154,10 @@ ENTRY(coprocessor_save) 1: l32i a0, a1, 0 retw +ENDPROC(coprocessor_save) + ENTRY(coprocessor_load) + entry a1, 32 s32i a0, a1, 0 movi a0, .Lload_cp_regs_jump_table @@ -162,8 +169,10 @@ ENTRY(coprocessor_load) 1: l32i a0, a1, 0 retw +ENDPROC(coprocessor_load) + /* - * coprocessor_flush(struct task_info*, index) + * coprocessor_flush(struct task_info*, index) * a2 a3 * coprocessor_restore(struct task_info*, index) * a2 a3 @@ -178,6 +187,7 @@ ENTRY(coprocessor_load) ENTRY(coprocessor_flush) + entry a1, 32 s32i a0, a1, 0 movi a0, .Lsave_cp_regs_jump_table @@ -191,6 +201,8 @@ ENTRY(coprocessor_flush) 1: l32i a0, a1, 0 retw +ENDPROC(coprocessor_flush) + ENTRY(coprocessor_restore) entry a1, 32 s32i a0, a1, 0 @@ -205,37 +217,40 @@ ENTRY(coprocessor_restore) 1: l32i a0, a1, 0 retw +ENDPROC(coprocessor_restore) + /* * Entry condition: * * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception */ ENTRY(fast_coprocessor_double) - wsr a0, EXCSAVE_1 + + wsr a0, excsave1 movi a0, unrecoverable_exception callx0 a0 +ENDPROC(fast_coprocessor_double) ENTRY(fast_coprocessor) /* Save remaining registers a1-a3 and SAR */ - xsr a3, EXCSAVE_1 s32i a3, a2, PT_AREG3 - rsr a3, SAR + rsr a3, sar s32i a1, a2, PT_AREG1 s32i a3, a2, PT_SAR mov a1, a2 - rsr a2, DEPC + rsr a2, depc s32i a2, a1, PT_AREG2 /* @@ -248,17 +263,17 @@ ENTRY(fast_coprocessor) /* Find coprocessor number. Subtract first CP EXCCAUSE from EXCCAUSE */ - rsr a3, EXCCAUSE + rsr a3, exccause addi a3, a3, -EXCCAUSE_COPROCESSOR0_DISABLED /* Set corresponding CPENABLE bit -> (sar:cp-index, a3: 1<<cp-index)*/ ssl a3 # SAR: 32 - coprocessor_number movi a2, 1 - rsr a0, CPENABLE + rsr a0, cpenable sll a2, a2 or a0, a0, a2 - wsr a0, CPENABLE + wsr a0, cpenable rsync /* Retrieve previous owner. (a3 still holds CP number) */ @@ -291,7 +306,7 @@ ENTRY(fast_coprocessor) /* Note that only a0 and a1 were preserved. */ -2: rsr a3, EXCCAUSE +2: rsr a3, exccause addi a3, a3, -EXCCAUSE_COPROCESSOR0_DISABLED movi a0, coprocessor_owner addx4 a0, a3, a0 @@ -321,15 +336,20 @@ ENTRY(fast_coprocessor) l32i a0, a1, PT_SAR l32i a3, a1, PT_AREG3 l32i a2, a1, PT_AREG2 - wsr a0, SAR + wsr a0, sar l32i a0, a1, PT_AREG0 l32i a1, a1, PT_AREG1 rfe +ENDPROC(fast_coprocessor) + .data + ENTRY(coprocessor_owner) + .fill XCHAL_CP_MAX, 4, 0 -#endif /* XTENSA_HAVE_COPROCESSORS */ +END(coprocessor_owner) +#endif /* XTENSA_HAVE_COPROCESSORS */ diff --git a/arch/xtensa/kernel/entry.S b/arch/xtensa/kernel/entry.S index 7e623607339..ef7f4990722 100644 --- a/arch/xtensa/kernel/entry.S +++ b/arch/xtensa/kernel/entry.S @@ -7,7 +7,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2004-2007 by Tensilica Inc. + * Copyright (C) 2004 - 2008 by Tensilica Inc. * * Chris Zankel <chris@zankel.net> * @@ -31,8 +31,6 @@ /* Unimplemented features. */ #undef KERNEL_STACK_OVERFLOW_CHECK -#undef PREEMPTIBLE_KERNEL -#undef ALLOCA_EXCEPTION_IN_IRAM /* Not well tested. * @@ -92,9 +90,9 @@ * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original value in depc - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave1: a3 + * excsave1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -110,10 +108,9 @@ ENTRY(user_exception) - /* Save a2, a3, and depc, restore excsave_1 and set SP. */ + /* Save a1, a2, a3, and set SP. */ - xsr a3, EXCSAVE_1 - rsr a0, DEPC + rsr a0, depc s32i a1, a2, PT_AREG1 s32i a0, a2, PT_AREG2 s32i a3, a2, PT_AREG3 @@ -125,16 +122,21 @@ _user_exception: /* Save SAR and turn off single stepping */ movi a2, 0 - rsr a3, SAR - xsr a2, ICOUNTLEVEL + rsr a3, sar + xsr a2, icountlevel s32i a3, a1, PT_SAR s32i a2, a1, PT_ICOUNTLEVEL +#if XCHAL_HAVE_THREADPTR + rur a2, threadptr + s32i a2, a1, PT_THREADPTR +#endif + /* Rotate ws so that the current windowbase is at bit0. */ /* Assume ws = xxwww1yyyy. Rotate ws right, so that a2 = yyyyxxwww1 */ - rsr a2, WINDOWBASE - rsr a3, WINDOWSTART + rsr a2, windowbase + rsr a3, windowstart ssr a2 s32i a2, a1, PT_WINDOWBASE s32i a3, a1, PT_WINDOWSTART @@ -205,12 +207,12 @@ _user_exception: /* WINDOWBASE still in SAR! */ - rsr a2, SAR # original WINDOWBASE + rsr a2, sar # original WINDOWBASE movi a3, 1 ssl a2 sll a3, a3 - wsr a3, WINDOWSTART # set corresponding WINDOWSTART bit - wsr a2, WINDOWBASE # and WINDOWSTART + wsr a3, windowstart # set corresponding WINDOWSTART bit + wsr a2, windowbase # and WINDOWSTART rsync /* We are back to the original stack pointer (a1) */ @@ -219,6 +221,7 @@ _user_exception: j common_exception +ENDPROC(user_exception) /* * First-level exit handler for kernel exceptions @@ -232,9 +235,9 @@ _user_exception: * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -250,10 +253,9 @@ _user_exception: ENTRY(kernel_exception) - /* Save a0, a2, a3, DEPC and set SP. */ + /* Save a1, a2, a3, and set SP. */ - xsr a3, EXCSAVE_1 # restore a3, excsave_1 - rsr a0, DEPC # get a2 + rsr a0, depc # get a2 s32i a1, a2, PT_AREG1 s32i a0, a2, PT_AREG2 s32i a3, a2, PT_AREG3 @@ -265,16 +267,16 @@ _kernel_exception: /* Save SAR and turn off single stepping */ movi a2, 0 - rsr a3, SAR - xsr a2, ICOUNTLEVEL + rsr a3, sar + xsr a2, icountlevel s32i a3, a1, PT_SAR s32i a2, a1, PT_ICOUNTLEVEL /* Rotate ws so that the current windowbase is at bit0. */ /* Assume ws = xxwww1yyyy. Rotate ws right, so that a2 = yyyyxxwww1 */ - rsr a2, WINDOWBASE # don't need to save these, we only - rsr a3, WINDOWSTART # need shifted windowstart: windowmask + rsr a2, windowbase # don't need to save these, we only + rsr a3, windowstart # need shifted windowstart: windowmask ssr a2 slli a2, a3, 32-WSBITS src a2, a3, a2 @@ -323,24 +325,24 @@ common_exception: /* Save some registers, disable loops and clear the syscall flag. */ - rsr a2, DEBUGCAUSE - rsr a3, EPC_1 + rsr a2, debugcause + rsr a3, epc1 s32i a2, a1, PT_DEBUGCAUSE s32i a3, a1, PT_PC movi a2, -1 - rsr a3, EXCVADDR + rsr a3, excvaddr s32i a2, a1, PT_SYSCALL movi a2, 0 s32i a3, a1, PT_EXCVADDR - xsr a2, LCOUNT + xsr a2, lcount s32i a2, a1, PT_LCOUNT /* It is now save to restore the EXC_TABLE_FIXUP variable. */ - rsr a0, EXCCAUSE + rsr a0, exccause movi a3, 0 - rsr a2, EXCSAVE_1 + rsr a2, excsave1 s32i a0, a1, PT_EXCCAUSE s32i a3, a2, EXC_TABLE_FIXUP @@ -348,38 +350,62 @@ common_exception: * so we can allow exceptions and interrupts (*) again. * Set PS(EXCM = 0, UM = 0, RING = 0, OWB = 0, WOE = 1, INTLEVEL = X) * - * (*) We only allow interrupts if PS.INTLEVEL was not set to 1 before - * (interrupts disabled) and if this exception is not an interrupt. + * (*) We only allow interrupts if they were previously enabled and + * we're not handling an IRQ */ - rsr a3, PS - addi a0, a0, -4 - movi a2, 1 - extui a3, a3, 0, 1 # a3 = PS.INTLEVEL[0] - moveqz a3, a2, a0 # a3 = 1 iff interrupt exception + rsr a3, ps + addi a0, a0, -EXCCAUSE_LEVEL1_INTERRUPT + movi a2, LOCKLEVEL + extui a3, a3, PS_INTLEVEL_SHIFT, PS_INTLEVEL_WIDTH + # a3 = PS.INTLEVEL + moveqz a3, a2, a0 # a3 = LOCKLEVEL iff interrupt movi a2, 1 << PS_WOE_BIT or a3, a3, a2 - rsr a0, EXCCAUSE - xsr a3, PS + rsr a0, exccause + xsr a3, ps s32i a3, a1, PT_PS # save ps - /* Save LBEG, LEND */ + /* Save lbeg, lend */ - rsr a2, LBEG - rsr a3, LEND + rsr a2, lbeg + rsr a3, lend s32i a2, a1, PT_LBEG s32i a3, a1, PT_LEND + /* Save SCOMPARE1 */ + +#if XCHAL_HAVE_S32C1I + rsr a2, scompare1 + s32i a2, a1, PT_SCOMPARE1 +#endif + /* Save optional registers. */ save_xtregs_opt a1 a2 a4 a5 a6 a7 PT_XTREGS_OPT +#ifdef CONFIG_TRACE_IRQFLAGS + l32i a4, a1, PT_DEPC + /* Double exception means we came here with an exception + * while PS.EXCM was set, i.e. interrupts disabled. + */ + bgeui a4, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f + l32i a4, a1, PT_EXCCAUSE + bnei a4, EXCCAUSE_LEVEL1_INTERRUPT, 1f + /* We came here with an interrupt means interrupts were enabled + * and we've just disabled them. + */ + movi a4, trace_hardirqs_off + callx4 a4 +1: +#endif + /* Go to second-level dispatcher. Set up parameters to pass to the * exception handler and call the exception handler. */ - movi a4, exc_table + rsr a4, excsave1 mov a6, a1 # pass stack frame mov a7, a0 # pass EXCCAUSE addx4 a4, a0, a4 @@ -390,13 +416,18 @@ common_exception: callx4 a4 /* Jump here for exception exit */ - + .global common_exception_return common_exception_return: +1: + rsil a2, LOCKLEVEL + /* Jump if we are returning from kernel exceptions. */ -1: l32i a3, a1, PT_PS - _bbci.l a3, PS_UM_BIT, 4f + l32i a3, a1, PT_PS + GET_THREAD_INFO(a2, a1) + l32i a4, a2, TI_FLAGS + _bbci.l a3, PS_UM_BIT, 6f /* Specific to a user exception exit: * We need to check some flags for signal handling and rescheduling, @@ -405,18 +436,16 @@ common_exception_return: * Note that we don't disable interrupts here. */ - GET_THREAD_INFO(a2,a1) - l32i a4, a2, TI_FLAGS - _bbsi.l a4, TIF_NEED_RESCHED, 3f _bbsi.l a4, TIF_NOTIFY_RESUME, 2f - _bbci.l a4, TIF_SIGPENDING, 4f + _bbci.l a4, TIF_SIGPENDING, 5f 2: l32i a4, a1, PT_DEPC bgeui a4, VALID_DOUBLE_EXCEPTION_ADDRESS, 4f /* Call do_signal() */ + rsil a2, 0 movi a4, do_notify_resume # int do_notify_resume(struct pt_regs*) mov a6, a1 callx4 a4 @@ -424,15 +453,59 @@ common_exception_return: 3: /* Reschedule */ + rsil a2, 0 movi a4, schedule # void schedule (void) callx4 a4 j 1b -4: /* Restore optional registers. */ +#ifdef CONFIG_PREEMPT +6: + _bbci.l a4, TIF_NEED_RESCHED, 4f + + /* Check current_thread_info->preempt_count */ + + l32i a4, a2, TI_PRE_COUNT + bnez a4, 4f + movi a4, preempt_schedule_irq + callx4 a4 + j 1b +#endif + +5: +#ifdef CONFIG_DEBUG_TLB_SANITY + l32i a4, a1, PT_DEPC + bgeui a4, VALID_DOUBLE_EXCEPTION_ADDRESS, 4f + movi a4, check_tlb_sanity + callx4 a4 +#endif +6: +4: +#ifdef CONFIG_TRACE_IRQFLAGS + l32i a4, a1, PT_DEPC + /* Double exception means we came here with an exception + * while PS.EXCM was set, i.e. interrupts disabled. + */ + bgeui a4, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f + l32i a4, a1, PT_EXCCAUSE + bnei a4, EXCCAUSE_LEVEL1_INTERRUPT, 1f + /* We came here with an interrupt means interrupts were enabled + * and we'll reenable them on return. + */ + movi a4, trace_hardirqs_on + callx4 a4 +1: +#endif + /* Restore optional registers. */ load_xtregs_opt a1 a2 a4 a5 a6 a7 PT_XTREGS_OPT - wsr a3, PS /* disable interrupts */ + /* Restore SCOMPARE1 */ + +#if XCHAL_HAVE_S32C1I + l32i a2, a1, PT_SCOMPARE1 + wsr a2, scompare1 +#endif + wsr a3, ps /* disable interrupts */ _bbci.l a3, PS_UM_BIT, kernel_exception_exit @@ -444,12 +517,12 @@ user_exception_exit: l32i a2, a1, PT_WINDOWBASE l32i a3, a1, PT_WINDOWSTART - wsr a1, DEPC # use DEPC as temp storage - wsr a3, WINDOWSTART # restore WINDOWSTART + wsr a1, depc # use DEPC as temp storage + wsr a3, windowstart # restore WINDOWSTART ssr a2 # preserve user's WB in the SAR - wsr a2, WINDOWBASE # switch to user's saved WB + wsr a2, windowbase # switch to user's saved WB rsync - rsr a1, DEPC # restore stack pointer + rsr a1, depc # restore stack pointer l32i a2, a1, PT_WMASK # register frames saved (in bits 4...9) rotw -1 # we restore a4..a7 _bltui a6, 16, 1f # only have to restore current window? @@ -475,8 +548,8 @@ user_exception_exit: /* Clear unrestored registers (don't leak anything to user-land */ -1: rsr a0, WINDOWBASE - rsr a3, SAR +1: rsr a0, windowbase + rsr a3, sar sub a3, a0, a3 beqz a3, 2f extui a3, a3, 0, WBBITS @@ -495,6 +568,11 @@ user_exception_exit: * (if we have restored WSBITS-1 frames). */ +#if XCHAL_HAVE_THREADPTR + l32i a3, a1, PT_THREADPTR + wur a3, threadptr +#endif + 2: j common_exception_exit /* This is the kernel exception exit. @@ -504,29 +582,6 @@ user_exception_exit: kernel_exception_exit: -#ifdef PREEMPTIBLE_KERNEL - -#ifdef CONFIG_PREEMPT - - /* - * Note: We've just returned from a call4, so we have - * at least 4 addt'l regs. - */ - - /* Check current_thread_info->preempt_count */ - - GET_THREAD_INFO(a2) - l32i a3, a2, TI_PREEMPT - bnez a3, 1f - - l32i a2, a2, TI_FLAGS - -1: - -#endif - -#endif - /* Check if we have to do a movsp. * * We only have to do a movsp if the previous window-frame has @@ -556,7 +611,7 @@ kernel_exception_exit: /* Test WINDOWSTART now. If spilled, do the movsp */ - rsr a3, WINDOWSTART + rsr a3, windowstart addi a0, a3, -1 and a3, a3, a0 _bnez a3, common_exception_exit @@ -604,24 +659,24 @@ common_exception_exit: 1: l32i a2, a1, PT_PC l32i a3, a1, PT_SAR - wsr a2, EPC_1 - wsr a3, SAR + wsr a2, epc1 + wsr a3, sar /* Restore LBEG, LEND, LCOUNT */ l32i a2, a1, PT_LBEG l32i a3, a1, PT_LEND - wsr a2, LBEG + wsr a2, lbeg l32i a2, a1, PT_LCOUNT - wsr a3, LEND - wsr a2, LCOUNT + wsr a3, lend + wsr a2, lcount /* We control single stepping through the ICOUNTLEVEL register. */ l32i a2, a1, PT_ICOUNTLEVEL movi a3, -2 - wsr a2, ICOUNTLEVEL - wsr a3, ICOUNT + wsr a2, icountlevel + wsr a3, icount /* Check if it was double exception. */ @@ -636,11 +691,13 @@ common_exception_exit: l32i a1, a1, PT_AREG1 rfe -1: wsr a0, DEPC +1: wsr a0, depc l32i a0, a1, PT_AREG0 l32i a1, a1, PT_AREG1 rfde +ENDPROC(kernel_exception) + /* * Debug exception handler. * @@ -651,25 +708,25 @@ common_exception_exit: ENTRY(debug_exception) - rsr a0, EPS + XCHAL_DEBUGLEVEL + rsr a0, SREG_EPS + XCHAL_DEBUGLEVEL bbsi.l a0, PS_EXCM_BIT, 1f # exception mode - /* Set EPC_1 and EXCCAUSE */ + /* Set EPC1 and EXCCAUSE */ - wsr a2, DEPC # save a2 temporarily - rsr a2, EPC + XCHAL_DEBUGLEVEL - wsr a2, EPC_1 + wsr a2, depc # save a2 temporarily + rsr a2, SREG_EPC + XCHAL_DEBUGLEVEL + wsr a2, epc1 movi a2, EXCCAUSE_MAPPED_DEBUG - wsr a2, EXCCAUSE + wsr a2, exccause /* Restore PS to the value before the debug exc but with PS.EXCM set.*/ movi a2, 1 << PS_EXCM_BIT or a2, a0, a2 movi a0, debug_exception # restore a3, debug jump vector - wsr a2, PS - xsr a0, EXCSAVE + XCHAL_DEBUGLEVEL + wsr a2, ps + xsr a0, SREG_EXCSAVE + XCHAL_DEBUGLEVEL /* Switch to kernel/user stack, restore jump vector, and save a0 */ @@ -680,19 +737,19 @@ ENTRY(debug_exception) movi a0, 0 s32i a1, a2, PT_AREG1 s32i a0, a2, PT_DEPC # mark it as a regular exception - xsr a0, DEPC + xsr a0, depc s32i a3, a2, PT_AREG3 s32i a0, a2, PT_AREG2 mov a1, a2 j _kernel_exception -2: rsr a2, EXCSAVE_1 +2: rsr a2, excsave1 l32i a2, a2, EXC_TABLE_KSTK # load kernel stack pointer s32i a0, a2, PT_AREG0 movi a0, 0 s32i a1, a2, PT_AREG1 s32i a0, a2, PT_DEPC - xsr a0, DEPC + xsr a0, depc s32i a3, a2, PT_AREG3 s32i a0, a2, PT_AREG2 mov a1, a2 @@ -701,6 +758,7 @@ ENTRY(debug_exception) /* Debug exception while in exception mode. */ 1: j 1b // FIXME!! +ENDPROC(debug_exception) /* * We get here in case of an unrecoverable exception. @@ -732,12 +790,12 @@ ENTRY(unrecoverable_exception) movi a0, 1 movi a1, 0 - wsr a0, WINDOWSTART - wsr a1, WINDOWBASE + wsr a0, windowstart + wsr a1, windowbase rsync - movi a1, (1 << PS_WOE_BIT) | 1 - wsr a1, PS + movi a1, (1 << PS_WOE_BIT) | LOCKLEVEL + wsr a1, ps rsync movi a1, init_task @@ -751,6 +809,7 @@ ENTRY(unrecoverable_exception) 1: j 1b +ENDPROC(unrecoverable_exception) /* -------------------------- FAST EXCEPTION HANDLERS ----------------------- */ @@ -759,176 +818,64 @@ ENTRY(unrecoverable_exception) * * The ALLOCA handler is entered when user code executes the MOVSP * instruction and the caller's frame is not in the register file. - * In this case, the caller frame's a0..a3 are on the stack just - * below sp (a1), and this handler moves them. * - * For "MOVSP <ar>,<as>" without destination register a1, this routine - * simply moves the value from <as> to <ar> without moving the save area. + * This algorithm was taken from the Ross Morley's RTOS Porting Layer: + * + * /home/ross/rtos/porting/XtensaRTOS-PortingLayer-20090507/xtensa_vectors.S + * + * It leverages the existing window spill/fill routines and their support for + * double exceptions. The 'movsp' instruction will only cause an exception if + * the next window needs to be loaded. In fact this ALLOCA exception may be + * replaced at some point by changing the hardware to do a underflow exception + * of the proper size instead. + * + * This algorithm simply backs out the register changes started by the user + * excpetion handler, makes it appear that we have started a window underflow + * by rotating the window back and then setting the old window base (OWB) in + * the 'ps' register with the rolled back window base. The 'movsp' instruction + * will be re-executed and this time since the next window frames is in the + * active AR registers it won't cause an exception. + * + * If the WindowUnderflow code gets a TLB miss the page will get mapped + * the the partial windeowUnderflow will be handeled in the double exception + * handler. * * Entry condition: * * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception */ -#if XCHAL_HAVE_BE -#define _EXTUI_MOVSP_SRC(ar) extui ar, ar, 4, 4 -#define _EXTUI_MOVSP_DST(ar) extui ar, ar, 0, 4 -#else -#define _EXTUI_MOVSP_SRC(ar) extui ar, ar, 0, 4 -#define _EXTUI_MOVSP_DST(ar) extui ar, ar, 4, 4 -#endif - ENTRY(fast_alloca) + rsr a0, windowbase + rotw -1 + rsr a2, ps + extui a3, a2, PS_OWB_SHIFT, PS_OWB_WIDTH + xor a3, a3, a4 + l32i a4, a6, PT_AREG0 + l32i a1, a6, PT_DEPC + rsr a6, depc + wsr a1, depc + slli a3, a3, PS_OWB_SHIFT + xor a2, a2, a3 + wsr a2, ps + rsync - /* We shouldn't be in a double exception. */ - - l32i a0, a2, PT_DEPC - _bgeui a0, VALID_DOUBLE_EXCEPTION_ADDRESS, .Lunhandled_double - - rsr a0, DEPC # get a2 - s32i a4, a2, PT_AREG4 # save a4 and - s32i a0, a2, PT_AREG2 # a2 to stack - - /* Exit critical section. */ - - movi a0, 0 - s32i a0, a3, EXC_TABLE_FIXUP - - /* Restore a3, excsave_1 */ - - xsr a3, EXCSAVE_1 # make sure excsave_1 is valid for dbl. - rsr a4, EPC_1 # get exception address - s32i a3, a2, PT_AREG3 # save a3 to stack - -#ifdef ALLOCA_EXCEPTION_IN_IRAM -#error iram not supported -#else - /* Note: l8ui not allowed in IRAM/IROM!! */ - l8ui a0, a4, 1 # read as(src) from MOVSP instruction -#endif - movi a3, .Lmovsp_src - _EXTUI_MOVSP_SRC(a0) # extract source register number - addx8 a3, a0, a3 - jx a3 - -.Lunhandled_double: - wsr a0, EXCSAVE_1 - movi a0, unrecoverable_exception - callx0 a0 - - .align 8 -.Lmovsp_src: - l32i a3, a2, PT_AREG0; _j 1f; .align 8 - mov a3, a1; _j 1f; .align 8 - l32i a3, a2, PT_AREG2; _j 1f; .align 8 - l32i a3, a2, PT_AREG3; _j 1f; .align 8 - l32i a3, a2, PT_AREG4; _j 1f; .align 8 - mov a3, a5; _j 1f; .align 8 - mov a3, a6; _j 1f; .align 8 - mov a3, a7; _j 1f; .align 8 - mov a3, a8; _j 1f; .align 8 - mov a3, a9; _j 1f; .align 8 - mov a3, a10; _j 1f; .align 8 - mov a3, a11; _j 1f; .align 8 - mov a3, a12; _j 1f; .align 8 - mov a3, a13; _j 1f; .align 8 - mov a3, a14; _j 1f; .align 8 - mov a3, a15; _j 1f; .align 8 - -1: - -#ifdef ALLOCA_EXCEPTION_IN_IRAM -#error iram not supported -#else - l8ui a0, a4, 0 # read ar(dst) from MOVSP instruction -#endif - addi a4, a4, 3 # step over movsp - _EXTUI_MOVSP_DST(a0) # extract destination register - wsr a4, EPC_1 # save new epc_1 - - _bnei a0, 1, 1f # no 'movsp a1, ax': jump - - /* Move the save area. This implies the use of the L32E - * and S32E instructions, because this move must be done with - * the user's PS.RING privilege levels, not with ring 0 - * (kernel's) privileges currently active with PS.EXCM - * set. Note that we have stil registered a fixup routine with the - * double exception vector in case a double exception occurs. - */ - - /* a0,a4:avail a1:old user stack a2:exc. stack a3:new user stack. */ - - l32e a0, a1, -16 - l32e a4, a1, -12 - s32e a0, a3, -16 - s32e a4, a3, -12 - l32e a0, a1, -8 - l32e a4, a1, -4 - s32e a0, a3, -8 - s32e a4, a3, -4 - - /* Restore stack-pointer and all the other saved registers. */ - - mov a1, a3 - - l32i a4, a2, PT_AREG4 - l32i a3, a2, PT_AREG3 - l32i a0, a2, PT_AREG0 - l32i a2, a2, PT_AREG2 - rfe - - /* MOVSP <at>,<as> was invoked with <at> != a1. - * Because the stack pointer is not being modified, - * we should be able to just modify the pointer - * without moving any save area. - * The processor only traps these occurrences if the - * caller window isn't live, so unfortunately we can't - * use this as an alternate trap mechanism. - * So we just do the move. This requires that we - * resolve the destination register, not just the source, - * so there's some extra work. - * (PERHAPS NOT REALLY NEEDED, BUT CLEANER...) - */ - - /* a0 dst-reg, a1 user-stack, a2 stack, a3 value of src reg. */ - -1: movi a4, .Lmovsp_dst - addx8 a4, a0, a4 - jx a4 - - .align 8 -.Lmovsp_dst: - s32i a3, a2, PT_AREG0; _j 1f; .align 8 - mov a1, a3; _j 1f; .align 8 - s32i a3, a2, PT_AREG2; _j 1f; .align 8 - s32i a3, a2, PT_AREG3; _j 1f; .align 8 - s32i a3, a2, PT_AREG4; _j 1f; .align 8 - mov a5, a3; _j 1f; .align 8 - mov a6, a3; _j 1f; .align 8 - mov a7, a3; _j 1f; .align 8 - mov a8, a3; _j 1f; .align 8 - mov a9, a3; _j 1f; .align 8 - mov a10, a3; _j 1f; .align 8 - mov a11, a3; _j 1f; .align 8 - mov a12, a3; _j 1f; .align 8 - mov a13, a3; _j 1f; .align 8 - mov a14, a3; _j 1f; .align 8 - mov a15, a3; _j 1f; .align 8 - -1: l32i a4, a2, PT_AREG4 - l32i a3, a2, PT_AREG3 - l32i a0, a2, PT_AREG0 - l32i a2, a2, PT_AREG2 - rfe - + _bbci.l a4, 31, 4f + rotw -1 + _bbci.l a8, 30, 8f + rotw -1 + j _WindowUnderflow12 +8: j _WindowUnderflow8 +4: j _WindowUnderflow4 +ENDPROC(fast_alloca) /* * fast system calls. @@ -944,58 +891,61 @@ ENTRY(fast_alloca) * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table */ ENTRY(fast_syscall_kernel) /* Skip syscall. */ - rsr a0, EPC_1 + rsr a0, epc1 addi a0, a0, 3 - wsr a0, EPC_1 + wsr a0, epc1 l32i a0, a2, PT_DEPC bgeui a0, VALID_DOUBLE_EXCEPTION_ADDRESS, fast_syscall_unrecoverable - rsr a0, DEPC # get syscall-nr + rsr a0, depc # get syscall-nr _beqz a0, fast_syscall_spill_registers _beqi a0, __NR_xtensa, fast_syscall_xtensa j kernel_exception +ENDPROC(fast_syscall_kernel) + ENTRY(fast_syscall_user) /* Skip syscall. */ - rsr a0, EPC_1 + rsr a0, epc1 addi a0, a0, 3 - wsr a0, EPC_1 + wsr a0, epc1 l32i a0, a2, PT_DEPC bgeui a0, VALID_DOUBLE_EXCEPTION_ADDRESS, fast_syscall_unrecoverable - rsr a0, DEPC # get syscall-nr + rsr a0, depc # get syscall-nr _beqz a0, fast_syscall_spill_registers _beqi a0, __NR_xtensa, fast_syscall_xtensa j user_exception -ENTRY(fast_syscall_unrecoverable) +ENDPROC(fast_syscall_user) - /* Restore all states. */ +ENTRY(fast_syscall_unrecoverable) - l32i a0, a2, PT_AREG0 # restore a0 - xsr a2, DEPC # restore a2, depc - rsr a3, EXCSAVE_1 + /* Restore all states. */ - wsr a0, EXCSAVE_1 - movi a0, unrecoverable_exception - callx0 a0 + l32i a0, a2, PT_AREG0 # restore a0 + xsr a2, depc # restore a2, depc + wsr a0, excsave1 + movi a0, unrecoverable_exception + callx0 a0 +ENDPROC(fast_syscall_unrecoverable) /* * sysxtensa syscall handler @@ -1011,10 +961,10 @@ ENTRY(fast_syscall_unrecoverable) * a0: a2 (syscall-nr), original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in a0 and DEPC - * a3: dispatch table, original in excsave_1 + * a3: a3 * a4..a15: unchanged * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -1047,8 +997,6 @@ ENTRY(fast_syscall_unrecoverable) ENTRY(fast_syscall_xtensa) - xsr a3, EXCSAVE_1 # restore a3, excsave1 - s32i a7, a2, PT_AREG7 # we need an additional register movi a7, 4 # sizeof(unsigned int) access_ok a3, a7, a0, a2, .Leac # a0: scratch reg, a2: sp @@ -1101,7 +1049,7 @@ CATCH movi a2, -EINVAL rfe - +ENDPROC(fast_syscall_xtensa) /* fast_syscall_spill_registers. @@ -1111,9 +1059,9 @@ CATCH * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * Note: We assume the stack pointer is EXC_TABLE_KSTK in the fixup handler. */ @@ -1122,201 +1070,68 @@ ENTRY(fast_syscall_spill_registers) /* Register a FIXUP handler (pass current wb as a parameter) */ + xsr a3, excsave1 movi a0, fast_syscall_spill_registers_fixup s32i a0, a3, EXC_TABLE_FIXUP - rsr a0, WINDOWBASE + rsr a0, windowbase s32i a0, a3, EXC_TABLE_PARAM + xsr a3, excsave1 # restore a3 and excsave_1 - /* Save a3 and SAR on stack. */ + /* Save a3, a4 and SAR on stack. */ - rsr a0, SAR - xsr a3, EXCSAVE_1 # restore a3 and excsave_1 + rsr a0, sar s32i a3, a2, PT_AREG3 - s32i a4, a2, PT_AREG4 - s32i a0, a2, PT_AREG5 # store SAR to PT_AREG5 + s32i a0, a2, PT_SAR - /* The spill routine might clobber a7, a11, and a15. */ + /* The spill routine might clobber a4, a7, a8, a11, a12, and a15. */ + s32i a4, a2, PT_AREG4 s32i a7, a2, PT_AREG7 + s32i a8, a2, PT_AREG8 s32i a11, a2, PT_AREG11 + s32i a12, a2, PT_AREG12 s32i a15, a2, PT_AREG15 - call0 _spill_registers # destroys a3, a4, and SAR - - /* Advance PC, restore registers and SAR, and return from exception. */ - - l32i a3, a2, PT_AREG5 - l32i a4, a2, PT_AREG4 - l32i a0, a2, PT_AREG0 - wsr a3, SAR - l32i a3, a2, PT_AREG3 - - /* Restore clobbered registers. */ - - l32i a7, a2, PT_AREG7 - l32i a11, a2, PT_AREG11 - l32i a15, a2, PT_AREG15 - - movi a2, 0 - rfe - -/* Fixup handler. - * - * We get here if the spill routine causes an exception, e.g. tlb miss. - * We basically restore WINDOWBASE and WINDOWSTART to the condition when - * we entered the spill routine and jump to the user exception handler. - * - * a0: value of depc, original value in depc - * a2: trashed, original value in EXC_TABLE_DOUBLE_SAVE - * a3: exctable, original value in excsave1 - */ - -fast_syscall_spill_registers_fixup: - - rsr a2, WINDOWBASE # get current windowbase (a2 is saved) - xsr a0, DEPC # restore depc and a0 - ssl a2 # set shift (32 - WB) - - /* We need to make sure the current registers (a0-a3) are preserved. - * To do this, we simply set the bit for the current window frame - * in WS, so that the exception handlers save them to the task stack. - */ - - rsr a3, EXCSAVE_1 # get spill-mask - slli a2, a3, 1 # shift left by one - - slli a3, a2, 32-WSBITS - src a2, a2, a3 # a1 = xxwww1yyxxxwww1yy...... - wsr a2, WINDOWSTART # set corrected windowstart - - movi a3, exc_table - l32i a2, a3, EXC_TABLE_DOUBLE_SAVE # restore a2 - l32i a3, a3, EXC_TABLE_PARAM # original WB (in user task) - - /* Return to the original (user task) WINDOWBASE. - * We leave the following frame behind: - * a0, a1, a2 same - * a3: trashed (saved in excsave_1) - * depc: depc (we have to return to that address) - * excsave_1: a3 - */ - - wsr a3, WINDOWBASE - rsync - - /* We are now in the original frame when we entered _spill_registers: - * a0: return address - * a1: used, stack pointer - * a2: kernel stack pointer - * a3: available, saved in EXCSAVE_1 - * depc: exception address - * excsave: a3 - * Note: This frame might be the same as above. - */ - - /* Setup stack pointer. */ - - addi a2, a2, -PT_USER_SIZE - s32i a0, a2, PT_AREG0 - - /* Make sure we return to this fixup handler. */ - - movi a3, fast_syscall_spill_registers_fixup_return - s32i a3, a2, PT_DEPC # setup depc - - /* Jump to the exception handler. */ - - movi a3, exc_table - rsr a0, EXCCAUSE - addx4 a0, a0, a3 # find entry in table - l32i a0, a0, EXC_TABLE_FAST_USER # load handler - jx a0 - -fast_syscall_spill_registers_fixup_return: - - /* When we return here, all registers have been restored (a2: DEPC) */ - - wsr a2, DEPC # exception address - - /* Restore fixup handler. */ - - xsr a3, EXCSAVE_1 - movi a2, fast_syscall_spill_registers_fixup - s32i a2, a3, EXC_TABLE_FIXUP - rsr a2, WINDOWBASE - s32i a2, a3, EXC_TABLE_PARAM - l32i a2, a3, EXC_TABLE_KSTK - - /* Load WB at the time the exception occurred. */ - - rsr a3, SAR # WB is still in SAR - neg a3, a3 - wsr a3, WINDOWBASE - rsync - - /* Restore a3 and return. */ - - movi a3, exc_table - xsr a3, EXCSAVE_1 - - rfde - - -/* - * spill all registers. - * - * This is not a real function. The following conditions must be met: - * - * - must be called with call0. - * - uses a3, a4 and SAR. - * - the last 'valid' register of each frame are clobbered. - * - the caller must have registered a fixup handler - * (or be inside a critical section) - * - PS_EXCM must be set (PS_WOE cleared?) - */ - -ENTRY(_spill_registers) - /* * Rotate ws so that the current windowbase is at bit 0. * Assume ws = xxxwww1yy (www1 current window frame). * Rotate ws right so that a4 = yyxxxwww1. */ - rsr a4, WINDOWBASE - rsr a3, WINDOWSTART # a3 = xxxwww1yy - ssr a4 # holds WB - slli a4, a3, WSBITS - or a3, a3, a4 # a3 = xxxwww1yyxxxwww1yy + rsr a0, windowbase + rsr a3, windowstart # a3 = xxxwww1yy + ssr a0 # holds WB + slli a0, a3, WSBITS + or a3, a3, a0 # a3 = xxxwww1yyxxxwww1yy srl a3, a3 # a3 = 00xxxwww1yyxxxwww1 /* We are done if there are no more than the current register frame. */ extui a3, a3, 1, WSBITS-1 # a3 = 0yyxxxwww - movi a4, (1 << (WSBITS-1)) + movi a0, (1 << (WSBITS-1)) _beqz a3, .Lnospill # only one active frame? jump /* We want 1 at the top, so that we return to the current windowbase */ - or a3, a3, a4 # 1yyxxxwww + or a3, a3, a0 # 1yyxxxwww /* Skip empty frames - get 'oldest' WINDOWSTART-bit. */ - wsr a3, WINDOWSTART # save shifted windowstart - neg a4, a3 - and a3, a4, a3 # first bit set from right: 000010000 + wsr a3, windowstart # save shifted windowstart + neg a0, a3 + and a3, a0, a3 # first bit set from right: 000010000 - ffs_ws a4, a3 # a4: shifts to skip empty frames + ffs_ws a0, a3 # a0: shifts to skip empty frames movi a3, WSBITS - sub a4, a3, a4 # WSBITS-a4:number of 0-bits from right - ssr a4 # save in SAR for later. + sub a0, a3, a0 # WSBITS-a0:number of 0-bits from right + ssr a0 # save in SAR for later. - rsr a3, WINDOWBASE - add a3, a3, a4 - wsr a3, WINDOWBASE + rsr a3, windowbase + add a3, a3, a0 + wsr a3, windowbase rsync - rsr a3, WINDOWSTART + rsr a3, windowstart srl a3, a3 # shift windowstart /* WB is now just one frame below the oldest frame in the register @@ -1327,22 +1142,6 @@ ENTRY(_spill_registers) * we have to save 4,8. or 12 registers. */ - _bbsi.l a3, 1, .Lc4 - _bbsi.l a3, 2, .Lc8 - - /* Special case: we have a call12-frame starting at a4. */ - - _bbci.l a3, 3, .Lc12 # bit 3 shouldn't be zero! (Jump to Lc12 first) - - s32e a4, a1, -16 # a1 is valid with an empty spill area - l32e a4, a5, -12 - s32e a8, a4, -48 - mov a8, a4 - l32e a4, a1, -16 - j .Lc12c - -.Lnospill: - ret .Lloop: _bbsi.l a3, 1, .Lc4 _bbci.l a3, 2, .Lc12 @@ -1356,20 +1155,10 @@ ENTRY(_spill_registers) s32e a9, a4, -28 s32e a10, a4, -24 s32e a11, a4, -20 - srli a11, a3, 2 # shift windowbase by 2 rotw 2 _bnei a3, 1, .Lloop - -.Lexit: /* Done. Do the final rotation, set WS, and return. */ - - rotw 1 - rsr a3, WINDOWBASE - ssl a3 - movi a3, 1 - sll a3, a3 - wsr a3, WINDOWSTART - ret + j .Lexit .Lc4: s32e a4, a9, -16 s32e a5, a9, -12 @@ -1385,11 +1174,11 @@ ENTRY(_spill_registers) /* 12-register frame (call12) */ - l32e a2, a5, -12 - s32e a8, a2, -48 - mov a8, a2 + l32e a0, a5, -12 + s32e a8, a0, -48 + mov a8, a0 -.Lc12c: s32e a9, a8, -44 + s32e a9, a8, -44 s32e a10, a8, -40 s32e a11, a8, -36 s32e a12, a8, -32 @@ -1409,61 +1198,208 @@ ENTRY(_spill_registers) */ rotw 1 - mov a5, a13 + mov a4, a13 rotw -1 - s32e a4, a9, -16 - s32e a5, a9, -12 - s32e a6, a9, -8 - s32e a7, a9, -4 + s32e a4, a8, -16 + s32e a5, a8, -12 + s32e a6, a8, -8 + s32e a7, a8, -4 rotw 3 _beqi a3, 1, .Lexit j .Lloop -.Linvalid_mask: +.Lexit: - /* We get here because of an unrecoverable error in the window - * registers. If we are in user space, we kill the application, - * however, this condition is unrecoverable in kernel space. - */ + /* Done. Do the final rotation and set WS */ + + rotw 1 + rsr a3, windowbase + ssl a3 + movi a3, 1 + sll a3, a3 + wsr a3, windowstart +.Lnospill: - rsr a0, PS - _bbci.l a0, PS_UM_BIT, 1f + /* Advance PC, restore registers and SAR, and return from exception. */ + + l32i a3, a2, PT_SAR + l32i a0, a2, PT_AREG0 + wsr a3, sar + l32i a3, a2, PT_AREG3 + + /* Restore clobbered registers. */ - /* User space: Setup a dummy frame and kill application. + l32i a4, a2, PT_AREG4 + l32i a7, a2, PT_AREG7 + l32i a8, a2, PT_AREG8 + l32i a11, a2, PT_AREG11 + l32i a12, a2, PT_AREG12 + l32i a15, a2, PT_AREG15 + + movi a2, 0 + rfe + +.Linvalid_mask: + + /* We get here because of an unrecoverable error in the window + * registers, so set up a dummy frame and kill the user application. * Note: We assume EXC_TABLE_KSTK contains a valid stack pointer. */ movi a0, 1 movi a1, 0 - wsr a0, WINDOWSTART - wsr a1, WINDOWBASE + wsr a0, windowstart + wsr a1, windowbase rsync movi a0, 0 - movi a3, exc_table + rsr a3, excsave1 l32i a1, a3, EXC_TABLE_KSTK - wsr a3, EXCSAVE_1 - movi a4, (1 << PS_WOE_BIT) | 1 - wsr a4, PS + movi a4, (1 << PS_WOE_BIT) | LOCKLEVEL + wsr a4, ps rsync movi a6, SIGSEGV movi a4, do_exit callx4 a4 -1: /* Kernel space: PANIC! */ + /* shouldn't return, so panic */ - wsr a0, EXCSAVE_1 + wsr a0, excsave1 movi a0, unrecoverable_exception callx0 a0 # should not return 1: j 1b + +ENDPROC(fast_syscall_spill_registers) + +/* Fixup handler. + * + * We get here if the spill routine causes an exception, e.g. tlb miss. + * We basically restore WINDOWBASE and WINDOWSTART to the condition when + * we entered the spill routine and jump to the user exception handler. + * + * Note that we only need to restore the bits in windowstart that have not + * been spilled yet by the _spill_register routine. Luckily, a3 contains a + * rotated windowstart with only those bits set for frames that haven't been + * spilled yet. Because a3 is rotated such that bit 0 represents the register + * frame for the current windowbase - 1, we need to rotate a3 left by the + * value of the current windowbase + 1 and move it to windowstart. + * + * a0: value of depc, original value in depc + * a2: trashed, original value in EXC_TABLE_DOUBLE_SAVE + * a3: exctable, original value in excsave1 + */ + +ENTRY(fast_syscall_spill_registers_fixup) + + rsr a2, windowbase # get current windowbase (a2 is saved) + xsr a0, depc # restore depc and a0 + ssl a2 # set shift (32 - WB) + + /* We need to make sure the current registers (a0-a3) are preserved. + * To do this, we simply set the bit for the current window frame + * in WS, so that the exception handlers save them to the task stack. + * + * Note: we use a3 to set the windowbase, so we take a special care + * of it, saving it in the original _spill_registers frame across + * the exception handler call. + */ + + xsr a3, excsave1 # get spill-mask + slli a3, a3, 1 # shift left by one + addi a3, a3, 1 # set the bit for the current window frame + + slli a2, a3, 32-WSBITS + src a2, a3, a2 # a2 = xxwww1yyxxxwww1yy...... + wsr a2, windowstart # set corrected windowstart + + srli a3, a3, 1 + rsr a2, excsave1 + l32i a2, a2, EXC_TABLE_DOUBLE_SAVE # restore a2 + xsr a2, excsave1 + s32i a3, a2, EXC_TABLE_DOUBLE_SAVE # save a3 + l32i a3, a2, EXC_TABLE_PARAM # original WB (in user task) + xsr a2, excsave1 + + /* Return to the original (user task) WINDOWBASE. + * We leave the following frame behind: + * a0, a1, a2 same + * a3: trashed (saved in EXC_TABLE_DOUBLE_SAVE) + * depc: depc (we have to return to that address) + * excsave_1: exctable + */ + + wsr a3, windowbase + rsync + + /* We are now in the original frame when we entered _spill_registers: + * a0: return address + * a1: used, stack pointer + * a2: kernel stack pointer + * a3: available + * depc: exception address + * excsave: exctable + * Note: This frame might be the same as above. + */ + + /* Setup stack pointer. */ + + addi a2, a2, -PT_USER_SIZE + s32i a0, a2, PT_AREG0 + + /* Make sure we return to this fixup handler. */ + + movi a3, fast_syscall_spill_registers_fixup_return + s32i a3, a2, PT_DEPC # setup depc + + /* Jump to the exception handler. */ + + rsr a3, excsave1 + rsr a0, exccause + addx4 a0, a0, a3 # find entry in table + l32i a0, a0, EXC_TABLE_FAST_USER # load handler + l32i a3, a3, EXC_TABLE_DOUBLE_SAVE + jx a0 + +ENDPROC(fast_syscall_spill_registers_fixup) + +ENTRY(fast_syscall_spill_registers_fixup_return) + + /* When we return here, all registers have been restored (a2: DEPC) */ + + wsr a2, depc # exception address + + /* Restore fixup handler. */ + + rsr a2, excsave1 + s32i a3, a2, EXC_TABLE_DOUBLE_SAVE + movi a3, fast_syscall_spill_registers_fixup + s32i a3, a2, EXC_TABLE_FIXUP + rsr a3, windowbase + s32i a3, a2, EXC_TABLE_PARAM + l32i a2, a2, EXC_TABLE_KSTK + + /* Load WB at the time the exception occurred. */ + + rsr a3, sar # WB is still in SAR + neg a3, a3 + wsr a3, windowbase + rsync + + rsr a3, excsave1 + l32i a3, a3, EXC_TABLE_DOUBLE_SAVE + + rfde + +ENDPROC(fast_syscall_spill_registers_fixup_return) + #ifdef CONFIG_MMU /* * We should never get here. Bail out! @@ -1475,6 +1411,8 @@ ENTRY(fast_second_level_miss_double_kernel) callx0 a0 # should not return 1: j 1b +ENDPROC(fast_second_level_miss_double_kernel) + /* First-level entry handler for user, kernel, and double 2nd-level * TLB miss exceptions. Note that for now, user and kernel miss * exceptions share the same entry point and are handled identically. @@ -1487,9 +1425,9 @@ ENTRY(fast_second_level_miss_double_kernel) * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -1497,9 +1435,10 @@ ENTRY(fast_second_level_miss_double_kernel) ENTRY(fast_second_level_miss) - /* Save a1. Note: we don't expect a double exception. */ + /* Save a1 and a3. Note: we don't expect a double exception. */ s32i a1, a2, PT_AREG1 + s32i a3, a2, PT_AREG3 /* We need to map the page of PTEs for the user task. Find * the pointer to that page. Also, it's possible for tsk->mm @@ -1521,10 +1460,7 @@ ENTRY(fast_second_level_miss) l32i a0, a1, TASK_MM # tsk->mm beqz a0, 9f - - /* We deliberately destroy a3 that holds the exception table. */ - -8: rsr a3, EXCVADDR # fault address +8: rsr a3, excvaddr # fault address _PGD_OFFSET(a0, a3, a1) l32i a0, a0, 0 # read pmdval beqz a0, 2f @@ -1542,7 +1478,7 @@ ENTRY(fast_second_level_miss) * pteval = ((pmdval - PAGE_OFFSET) & PAGE_MASK) | PAGE_DIRECTORY */ - movi a1, -PAGE_OFFSET + movi a1, (-PAGE_OFFSET) & 0xffffffff add a0, a0, a1 # pmdval - PAGE_OFFSET extui a1, a0, 0, PAGE_SHIFT # ... & PAGE_MASK xor a0, a0, a1 @@ -1561,7 +1497,7 @@ ENTRY(fast_second_level_miss) */ extui a3, a3, 28, 2 # addr. bit 28 and 29 0,1,2,3 - rsr a1, PTEVADDR + rsr a1, ptevaddr addx2 a3, a3, a3 # -> 0,3,6,9 srli a1, a1, PAGE_SHIFT extui a3, a3, 2, 2 # -> 0,0,1,2 @@ -1574,7 +1510,7 @@ ENTRY(fast_second_level_miss) /* Exit critical section. */ -4: movi a3, exc_table # restore a3 +4: rsr a3, excsave1 movi a0, 0 s32i a0, a3, EXC_TABLE_FIXUP @@ -1582,19 +1518,19 @@ ENTRY(fast_second_level_miss) l32i a0, a2, PT_AREG0 l32i a1, a2, PT_AREG1 + l32i a3, a2, PT_AREG3 l32i a2, a2, PT_DEPC - xsr a3, EXCSAVE_1 bgeui a2, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f /* Restore excsave1 and return. */ - rsr a2, DEPC + rsr a2, depc rfe /* Return from double exception. */ -1: xsr a2, DEPC +1: xsr a2, depc esync rfde @@ -1618,7 +1554,7 @@ ENTRY(fast_second_level_miss) /* Make sure the exception originated in the special functions */ movi a0, __tlbtemp_mapping_start - rsr a3, EPC_1 + rsr a3, epc1 bltu a3, a0, 2f movi a0, __tlbtemp_mapping_end bgeu a3, a0, 2f @@ -1626,7 +1562,7 @@ ENTRY(fast_second_level_miss) /* Check if excvaddr was in one of the TLBTEMP_BASE areas. */ movi a3, TLBTEMP_BASE_1 - rsr a0, EXCVADDR + rsr a0, excvaddr bltu a0, a3, 2f addi a1, a0, -(2 << (DCACHE_ALIAS_ORDER + PAGE_SHIFT)) @@ -1635,7 +1571,7 @@ ENTRY(fast_second_level_miss) /* Check if we have to restore an ITLB mapping. */ movi a1, __tlbtemp_mapping_itlb - rsr a3, EPC_1 + rsr a3, epc1 sub a3, a3, a1 /* Calculate VPN */ @@ -1670,18 +1606,16 @@ ENTRY(fast_second_level_miss) 2: /* Invalid PGD, default exception handling */ - movi a3, exc_table - rsr a1, DEPC - xsr a3, EXCSAVE_1 + rsr a1, depc s32i a1, a2, PT_AREG2 - s32i a3, a2, PT_AREG3 mov a1, a2 - rsr a2, PS + rsr a2, ps bbsi.l a2, PS_UM_BIT, 1f j _kernel_exception 1: j _user_exception +ENDPROC(fast_second_level_miss) /* * StoreProhibitedException @@ -1693,9 +1627,9 @@ ENTRY(fast_second_level_miss) * a0: trashed, original value saved on stack (PT_AREG0) * a1: a1 * a2: new stack pointer, original in DEPC - * a3: dispatch table + * a3: a3 * depc: a2, original value saved on stack (PT_DEPC) - * excsave_1: a3 + * excsave_1: dispatch table * * PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC * < VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception @@ -1703,61 +1637,64 @@ ENTRY(fast_second_level_miss) ENTRY(fast_store_prohibited) - /* Save a1 and a4. */ + /* Save a1 and a3. */ s32i a1, a2, PT_AREG1 - s32i a4, a2, PT_AREG4 + s32i a3, a2, PT_AREG3 GET_CURRENT(a1,a2) l32i a0, a1, TASK_MM # tsk->mm beqz a0, 9f -8: rsr a1, EXCVADDR # fault address - _PGD_OFFSET(a0, a1, a4) +8: rsr a1, excvaddr # fault address + _PGD_OFFSET(a0, a1, a3) l32i a0, a0, 0 beqz a0, 2f - /* Note that we assume _PAGE_WRITABLE_BIT is only set if pte is valid.*/ + /* + * Note that we test _PAGE_WRITABLE_BIT only if PTE is present + * and is not PAGE_NONE. See pgtable.h for possible PTE layouts. + */ - _PTE_OFFSET(a0, a1, a4) - l32i a4, a0, 0 # read pteval - bbci.l a4, _PAGE_WRITABLE_BIT, 2f + _PTE_OFFSET(a0, a1, a3) + l32i a3, a0, 0 # read pteval + movi a1, _PAGE_CA_INVALID + ball a3, a1, 2f + bbci.l a3, _PAGE_WRITABLE_BIT, 2f movi a1, _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_HW_WRITE - or a4, a4, a1 - rsr a1, EXCVADDR - s32i a4, a0, 0 + or a3, a3, a1 + rsr a1, excvaddr + s32i a3, a0, 0 /* We need to flush the cache if we have page coloring. */ #if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK dhwb a0, 0 #endif pdtlb a0, a1 - wdtlb a4, a0 + wdtlb a3, a0 /* Exit critical section. */ movi a0, 0 + rsr a3, excsave1 s32i a0, a3, EXC_TABLE_FIXUP /* Restore the working registers, and return. */ - l32i a4, a2, PT_AREG4 + l32i a3, a2, PT_AREG3 l32i a1, a2, PT_AREG1 l32i a0, a2, PT_AREG0 l32i a2, a2, PT_DEPC - /* Restore excsave1 and a3. */ - - xsr a3, EXCSAVE_1 bgeui a2, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f - rsr a2, DEPC + rsr a2, depc rfe /* Double exception. Restore FIXUP handler and return. */ -1: xsr a2, DEPC +1: xsr a2, depc esync rfde @@ -1766,17 +1703,17 @@ ENTRY(fast_store_prohibited) 2: /* If there was a problem, handle fault in C */ - rsr a4, DEPC # still holds a2 - xsr a3, EXCSAVE_1 - s32i a4, a2, PT_AREG2 - s32i a3, a2, PT_AREG3 - l32i a4, a2, PT_AREG4 + rsr a3, depc # still holds a2 + s32i a3, a2, PT_AREG2 mov a1, a2 - rsr a2, PS + rsr a2, ps bbsi.l a2, PS_UM_BIT, 1f j _kernel_exception 1: j _user_exception + +ENDPROC(fast_store_prohibited) + #endif /* CONFIG_MMU */ /* @@ -1787,6 +1724,7 @@ ENTRY(fast_store_prohibited) */ ENTRY(system_call) + entry a1, 32 /* regs->syscall = regs->areg[2] */ @@ -1831,50 +1769,45 @@ ENTRY(system_call) callx4 a4 retw +ENDPROC(system_call) /* - * Create a kernel thread + * Spill live registers on the kernel stack macro. * - * int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) - * a2 a2 a3 a4 + * Entry condition: ps.woe is set, ps.excm is cleared + * Exit condition: windowstart has single bit set + * May clobber: a12, a13 */ + .macro spill_registers_kernel -ENTRY(kernel_thread) - entry a1, 16 - - mov a5, a2 # preserve fn over syscall - mov a7, a3 # preserve args over syscall - - movi a3, _CLONE_VM | _CLONE_UNTRACED - movi a2, __NR_clone - or a6, a4, a3 # arg0: flags - mov a3, a1 # arg1: sp - syscall - - beq a3, a1, 1f # branch if parent - mov a6, a7 # args - callx4 a5 # fn(args) - - movi a2, __NR_exit - syscall # return value of fn(args) still in a6 - -1: retw - -/* - * Do a system call from kernel instead of calling sys_execve, so we end up - * with proper pt_regs. - * - * int kernel_execve(const char *fname, char *const argv[], charg *const envp[]) - * a2 a2 a3 a4 - */ - -ENTRY(kernel_execve) - entry a1, 16 - mov a6, a2 # arg0 is in a6 - movi a2, __NR_execve - syscall - +#if XCHAL_NUM_AREGS > 16 + call12 1f + _j 2f + retw + .align 4 +1: + _entry a1, 48 + addi a12, a0, 3 +#if XCHAL_NUM_AREGS > 32 + .rept (XCHAL_NUM_AREGS - 32) / 12 + _entry a1, 48 + mov a12, a0 + .endr +#endif + _entry a1, 48 +#if XCHAL_NUM_AREGS % 12 == 0 + mov a8, a8 +#elif XCHAL_NUM_AREGS % 12 == 4 + mov a12, a12 +#elif XCHAL_NUM_AREGS % 12 == 8 + mov a4, a4 +#endif retw +2: +#else + mov a12, a12 +#endif + .endm /* * Task switch. @@ -1887,22 +1820,21 @@ ENTRY(_switch_to) entry a1, 16 - mov a12, a2 # preserve 'prev' (a2) - mov a13, a3 # and 'next' (a3) + mov a10, a2 # preserve 'prev' (a2) + mov a11, a3 # and 'next' (a3) l32i a4, a2, TASK_THREAD_INFO l32i a5, a3, TASK_THREAD_INFO - save_xtregs_user a4 a6 a8 a9 a10 a11 THREAD_XTREGS_USER + save_xtregs_user a4 a6 a8 a9 a12 a13 THREAD_XTREGS_USER - s32i a0, a12, THREAD_RA # save return address - s32i a1, a12, THREAD_SP # save stack pointer + s32i a0, a10, THREAD_RA # save return address + s32i a1, a10, THREAD_SP # save stack pointer /* Disable ints while we manipulate the stack pointer. */ - movi a14, (1 << PS_EXCM_BIT) | LOCKLEVEL - xsr a14, PS - rsr a3, EXCSAVE_1 + rsil a14, LOCKLEVEL + rsr a3, excsave1 rsync s32i a3, a3, EXC_TABLE_FIXUP /* enter critical section */ @@ -1910,13 +1842,13 @@ ENTRY(_switch_to) #if (XTENSA_HAVE_COPROCESSORS || XTENSA_HAVE_IO_PORTS) l32i a3, a5, THREAD_CPENABLE - xsr a3, CPENABLE + xsr a3, cpenable s32i a3, a4, THREAD_CPENABLE #endif /* Flush register file. */ - call0 _spill_registers # destroys a3, a4, and SAR + spill_registers_kernel /* Set kernel stack (and leave critical section) * Note: It's save to set it here. The stack will not be overwritten @@ -1924,25 +1856,26 @@ ENTRY(_switch_to) * we return from kernel space. */ - rsr a3, EXCSAVE_1 # exc_table + rsr a3, excsave1 # exc_table movi a6, 0 addi a7, a5, PT_REGS_OFFSET s32i a6, a3, EXC_TABLE_FIXUP s32i a7, a3, EXC_TABLE_KSTK - /* restore context of the task that 'next' addresses */ + /* restore context of the task 'next' */ - l32i a0, a13, THREAD_RA # restore return address - l32i a1, a13, THREAD_SP # restore stack pointer + l32i a0, a11, THREAD_RA # restore return address + l32i a1, a11, THREAD_SP # restore stack pointer - load_xtregs_user a5 a6 a8 a9 a10 a11 THREAD_XTREGS_USER + load_xtregs_user a5 a6 a8 a9 a12 a13 THREAD_XTREGS_USER - wsr a14, PS - mov a2, a12 # return 'prev' + wsr a14, ps + mov a2, a10 # return 'prev' rsync retw +ENDPROC(_switch_to) ENTRY(ret_from_fork) @@ -1958,3 +1891,18 @@ ENTRY(ret_from_fork) j common_exception_return +ENDPROC(ret_from_fork) + +/* + * Kernel thread creation helper + * On entry, set up by copy_thread: a2 = thread_fn, a3 = thread_fn arg + * left from _switch_to: a6 = prev + */ +ENTRY(ret_from_kernel_thread) + + call4 schedule_tail + mov a6, a3 + callx4 a2 + j common_exception_return + +ENDPROC(ret_from_kernel_thread) diff --git a/arch/xtensa/kernel/head.S b/arch/xtensa/kernel/head.S index 3ef91a73652..aeeb3cc8a41 100644 --- a/arch/xtensa/kernel/head.S +++ b/arch/xtensa/kernel/head.S @@ -7,7 +7,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2001 - 2005 Tensilica Inc. + * Copyright (C) 2001 - 2008 Tensilica Inc. * * Chris Zankel <chris@zankel.net> * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca> @@ -18,6 +18,8 @@ #include <asm/processor.h> #include <asm/page.h> #include <asm/cacheasm.h> +#include <asm/initialize_mmu.h> +#include <asm/mxregs.h> #include <linux/init.h> #include <linux/linkage.h> @@ -47,34 +49,65 @@ */ __HEAD - .globl _start -_start: _j 2f - .align 4 -1: .word _startup -2: l32r a0, 1b - jx a0 - - .section .init.text, "ax" - .align 4 -_startup: + .begin no-absolute-literals - /* Disable interrupts and exceptions. */ - - movi a0, LOCKLEVEL - wsr a0, PS +ENTRY(_start) /* Preserve the pointer to the boot parameter list in EXCSAVE_1 */ + wsr a2, excsave1 + _j _SetupOCD - wsr a2, EXCSAVE_1 - - /* Start with a fresh windowbase and windowstart. */ + .align 4 + .literal_position +.Lstartup: + .word _startup + .align 4 +_SetupOCD: + /* + * Initialize WB, WS, and clear PS.EXCM (to allow loop instructions). + * Set Interrupt Level just below XCHAL_DEBUGLEVEL to allow + * xt-gdb to single step via DEBUG exceptions received directly + * by ocd. + */ movi a1, 1 movi a0, 0 - wsr a1, WINDOWSTART - wsr a0, WINDOWBASE + wsr a1, windowstart + wsr a0, windowbase + rsync + + movi a1, LOCKLEVEL + wsr a1, ps rsync + .global _SetupMMU +_SetupMMU: + Offset = _SetupMMU - _start + +#ifdef CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX + initialize_mmu +#if defined(CONFIG_MMU) && XCHAL_HAVE_PTP_MMU && XCHAL_HAVE_SPANNING_WAY + rsr a2, excsave1 + movi a3, 0x08000000 + bgeu a2, a3, 1f + movi a3, 0xd0000000 + add a2, a2, a3 + wsr a2, excsave1 +1: +#endif +#endif + .end no-absolute-literals + + l32r a0, .Lstartup + jx a0 + +ENDPROC(_start) + + __REF + .literal_position + +ENTRY(_startup) + /* Set a0 to 0 for the remaining initialization. */ movi a0, 0 @@ -82,59 +115,50 @@ _startup: /* Clear debugging registers. */ #if XCHAL_HAVE_DEBUG - wsr a0, IBREAKENABLE - wsr a0, ICOUNT +#if XCHAL_NUM_IBREAK > 0 + wsr a0, ibreakenable +#endif + wsr a0, icount movi a1, 15 - wsr a0, ICOUNTLEVEL + wsr a0, icountlevel .set _index, 0 .rept XCHAL_NUM_DBREAK - 1 - wsr a0, DBREAKC + _index + wsr a0, SREG_DBREAKC + _index .set _index, _index + 1 .endr #endif /* Clear CCOUNT (not really necessary, but nice) */ - wsr a0, CCOUNT # not really necessary, but nice + wsr a0, ccount # not really necessary, but nice /* Disable zero-loops. */ #if XCHAL_HAVE_LOOPS - wsr a0, LCOUNT + wsr a0, lcount #endif /* Disable all timers. */ .set _index, 0 - .rept XCHAL_NUM_TIMERS - 1 - wsr a0, CCOMPARE + _index + .rept XCHAL_NUM_TIMERS + wsr a0, SREG_CCOMPARE + _index .set _index, _index + 1 .endr /* Interrupt initialization. */ movi a2, XCHAL_INTTYPE_MASK_SOFTWARE | XCHAL_INTTYPE_MASK_EXTERN_EDGE - wsr a0, INTENABLE - wsr a2, INTCLEAR + wsr a0, intenable + wsr a2, intclear /* Disable coprocessors. */ -#if XCHAL_CP_NUM > 0 - wsr a0, CPENABLE +#if XCHAL_HAVE_CP + wsr a0, cpenable #endif - /* Set PS.INTLEVEL=1, PS.WOE=0, kernel stack, PS.EXCM=0 - * - * Note: PS.EXCM must be cleared before using any loop - * instructions; otherwise, they are silently disabled, and - * at most one iteration of the loop is executed. - */ - - movi a1, 1 - wsr a1, PS - rsync - /* Initialize the caches. * a2, a3 are just working registers (clobbered). */ @@ -152,6 +176,37 @@ _startup: isync +#ifdef CONFIG_HAVE_SMP + movi a2, CCON # MX External Register to Configure Cache + movi a3, 1 + wer a3, a2 +#endif + + /* Setup stack and enable window exceptions (keep irqs disabled) */ + + movi a1, start_info + l32i a1, a1, 0 + + movi a2, (1 << PS_WOE_BIT) | LOCKLEVEL + # WOE=1, INTLEVEL=LOCKLEVEL, UM=0 + wsr a2, ps # (enable reg-windows; progmode stack) + rsync + + /* Set up EXCSAVE[DEBUGLEVEL] to point to the Debug Exception Handler.*/ + + movi a2, debug_exception + wsr a2, SREG_EXCSAVE + XCHAL_DEBUGLEVEL + +#ifdef CONFIG_SMP + /* + * Notice that we assume with SMP that cores have PRID + * supported by the cores. + */ + rsr a2, prid + bnez a2, .Lboot_secondary + +#endif /* CONFIG_SMP */ + /* Unpack data sections * * The linker script used to build the Linux kernel image @@ -199,25 +254,13 @@ _startup: ___flush_dcache_all a2 a3 #endif + memw + isync + ___invalidate_icache_all a2 a3 + isync - /* Setup stack and enable window exceptions (keep irqs disabled) */ - - movi a1, init_thread_union - addi a1, a1, KERNEL_STACK_SIZE - - movi a2, 0x00040001 # WOE=1, INTLEVEL=1, UM=0 - wsr a2, PS # (enable reg-windows; progmode stack) - rsync - - /* Set up EXCSAVE[DEBUGLEVEL] to point to the Debug Exception Handler.*/ - - movi a2, debug_exception - wsr a2, EXCSAVE + XCHAL_DEBUGLEVEL - - /* Set up EXCSAVE[1] to point to the exc_table. */ - - movi a6, exc_table - xsr a6, EXCSAVE_1 + movi a6, 0 + xsr a6, excsave1 /* init_arch kick-starts the linux kernel */ @@ -230,6 +273,92 @@ _startup: should_never_return: j should_never_return +#ifdef CONFIG_SMP +.Lboot_secondary: + + movi a2, cpu_start_ccount +1: + l32i a3, a2, 0 + beqi a3, 0, 1b + movi a3, 0 + s32i a3, a2, 0 + memw +1: + l32i a3, a2, 0 + beqi a3, 0, 1b + wsr a3, ccount + movi a3, 0 + s32i a3, a2, 0 + memw + + movi a6, 0 + wsr a6, excsave1 + + movi a4, secondary_start_kernel + callx4 a4 + j should_never_return + +#endif /* CONFIG_SMP */ + +ENDPROC(_startup) + +#ifdef CONFIG_HOTPLUG_CPU + +ENTRY(cpu_restart) + +#if XCHAL_DCACHE_IS_WRITEBACK + ___flush_invalidate_dcache_all a2 a3 +#else + ___invalidate_dcache_all a2 a3 +#endif + memw + movi a2, CCON # MX External Register to Configure Cache + movi a3, 0 + wer a3, a2 + extw + + rsr a0, prid + neg a2, a0 + movi a3, cpu_start_id + s32i a2, a3, 0 +#if XCHAL_DCACHE_IS_WRITEBACK + dhwbi a3, 0 +#endif +1: + l32i a2, a3, 0 + dhi a3, 0 + bne a2, a0, 1b + + /* + * Initialize WB, WS, and clear PS.EXCM (to allow loop instructions). + * Set Interrupt Level just below XCHAL_DEBUGLEVEL to allow + * xt-gdb to single step via DEBUG exceptions received directly + * by ocd. + */ + movi a1, 1 + movi a0, 0 + wsr a1, windowstart + wsr a0, windowbase + rsync + + movi a1, LOCKLEVEL + wsr a1, ps + rsync + + j _startup + +ENDPROC(cpu_restart) + +#endif /* CONFIG_HOTPLUG_CPU */ + +/* + * DATA section + */ + + .section ".data.init.refok" + .align 4 +ENTRY(start_info) + .long init_thread_union + KERNEL_STACK_SIZE /* * BSS section @@ -239,6 +368,8 @@ __PAGE_ALIGNED_BSS #ifdef CONFIG_MMU ENTRY(swapper_pg_dir) .fill PAGE_SIZE, 1, 0 +END(swapper_pg_dir) #endif ENTRY(empty_zero_page) .fill PAGE_SIZE, 1, 0 +END(empty_zero_page) diff --git a/arch/xtensa/kernel/io.c b/arch/xtensa/kernel/io.c deleted file mode 100644 index 5b65269b1d2..00000000000 --- a/arch/xtensa/kernel/io.c +++ /dev/null @@ -1,75 +0,0 @@ -/* - * arch/xtensa/io.c - * - * IO primitives - * - * This program is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License as published by the - * Free Software Foundation; either version 2 of the License, or (at your - * option) any later version. - * - * Copied from sparc. - * - * Chris Zankel <chris@zankel.net> - * - */ - -#include <asm/io.h> -#include <asm/byteorder.h> - -void outsb(unsigned long addr, const void *src, unsigned long count) { - while (count) { - count -= 1; - writeb(*(const char *)src, addr); - src += 1; - addr += 1; - } -} - -void outsw(unsigned long addr, const void *src, unsigned long count) { - while (count) { - count -= 2; - writew(*(const short *)src, addr); - src += 2; - addr += 2; - } -} - -void outsl(unsigned long addr, const void *src, unsigned long count) { - while (count) { - count -= 4; - writel(*(const long *)src, addr); - src += 4; - addr += 4; - } -} - -void insb(unsigned long addr, void *dst, unsigned long count) { - while (count) { - count -= 1; - *(unsigned char *)dst = readb(addr); - dst += 1; - addr += 1; - } -} - -void insw(unsigned long addr, void *dst, unsigned long count) { - while (count) { - count -= 2; - *(unsigned short *)dst = readw(addr); - dst += 2; - addr += 2; - } -} - -void insl(unsigned long addr, void *dst, unsigned long count) { - while (count) { - count -= 4; - /* - * XXX I am sure we are in for an unaligned trap here. - */ - *(unsigned long *)dst = readl(addr); - dst += 4; - addr += 4; - } -} diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c index 4340ee076bd..3eee94f621e 100644 --- a/arch/xtensa/kernel/irq.c +++ b/arch/xtensa/kernel/irq.c @@ -4,7 +4,7 @@ * Xtensa built-in interrupt controller and some generic functions copied * from i386. * - * Copyright (C) 2002 - 2006 Tensilica, Inc. + * Copyright (C) 2002 - 2013 Tensilica, Inc. * Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar * * @@ -18,31 +18,27 @@ #include <linux/interrupt.h> #include <linux/irq.h> #include <linux/kernel_stat.h> +#include <linux/irqchip.h> +#include <linux/irqchip/xtensa-mx.h> +#include <linux/irqchip/xtensa-pic.h> +#include <linux/irqdomain.h> +#include <linux/of.h> +#include <asm/mxregs.h> #include <asm/uaccess.h> #include <asm/platform.h> -static unsigned int cached_irq_mask; - atomic_t irq_err_count; -/* - * do_IRQ handles all normal device IRQ's (the special - * SMP cross-CPU interrupts have their own specific - * handlers). - */ - -asmlinkage void do_IRQ(int irq, struct pt_regs *regs) +asmlinkage void do_IRQ(int hwirq, struct pt_regs *regs) { - struct pt_regs *old_regs = set_irq_regs(regs); + int irq = irq_find_mapping(NULL, hwirq); - if (irq >= NR_IRQS) { + if (hwirq >= NR_IRQS) { printk(KERN_EMERG "%s: cannot handle IRQ %d\n", - __func__, irq); + __func__, hwirq); } - irq_enter(); - #ifdef CONFIG_DEBUG_STACKOVERFLOW /* Debugging check for stack overflow: is there less than 1KB free? */ { @@ -57,95 +53,136 @@ asmlinkage void do_IRQ(int irq, struct pt_regs *regs) } #endif generic_handle_irq(irq); - - irq_exit(); - set_irq_regs(old_regs); } int arch_show_interrupts(struct seq_file *p, int prec) { +#ifdef CONFIG_SMP + show_ipi_list(p, prec); +#endif seq_printf(p, "%*s: ", prec, "ERR"); seq_printf(p, "%10u\n", atomic_read(&irq_err_count)); return 0; } -static void xtensa_irq_mask(struct irq_data *d) +int xtensa_irq_domain_xlate(const u32 *intspec, unsigned int intsize, + unsigned long int_irq, unsigned long ext_irq, + unsigned long *out_hwirq, unsigned int *out_type) { - cached_irq_mask &= ~(1 << d->irq); - set_sr (cached_irq_mask, INTENABLE); + if (WARN_ON(intsize < 1 || intsize > 2)) + return -EINVAL; + if (intsize == 2 && intspec[1] == 1) { + int_irq = xtensa_map_ext_irq(ext_irq); + if (int_irq < XCHAL_NUM_INTERRUPTS) + *out_hwirq = int_irq; + else + return -EINVAL; + } else { + *out_hwirq = int_irq; + } + *out_type = IRQ_TYPE_NONE; + return 0; } -static void xtensa_irq_unmask(struct irq_data *d) +int xtensa_irq_map(struct irq_domain *d, unsigned int irq, + irq_hw_number_t hw) { - cached_irq_mask |= 1 << d->irq; - set_sr (cached_irq_mask, INTENABLE); + struct irq_chip *irq_chip = d->host_data; + u32 mask = 1 << hw; + + if (mask & XCHAL_INTTYPE_MASK_SOFTWARE) { + irq_set_chip_and_handler_name(irq, irq_chip, + handle_simple_irq, "level"); + irq_set_status_flags(irq, IRQ_LEVEL); + } else if (mask & XCHAL_INTTYPE_MASK_EXTERN_EDGE) { + irq_set_chip_and_handler_name(irq, irq_chip, + handle_edge_irq, "edge"); + irq_clear_status_flags(irq, IRQ_LEVEL); + } else if (mask & XCHAL_INTTYPE_MASK_EXTERN_LEVEL) { + irq_set_chip_and_handler_name(irq, irq_chip, + handle_level_irq, "level"); + irq_set_status_flags(irq, IRQ_LEVEL); + } else if (mask & XCHAL_INTTYPE_MASK_TIMER) { + irq_set_chip_and_handler_name(irq, irq_chip, + handle_percpu_irq, "timer"); + irq_clear_status_flags(irq, IRQ_LEVEL); + } else {/* XCHAL_INTTYPE_MASK_WRITE_ERROR */ + /* XCHAL_INTTYPE_MASK_NMI */ + irq_set_chip_and_handler_name(irq, irq_chip, + handle_level_irq, "level"); + irq_set_status_flags(irq, IRQ_LEVEL); + } + return 0; } -static void xtensa_irq_enable(struct irq_data *d) +unsigned xtensa_map_ext_irq(unsigned ext_irq) { - variant_irq_enable(d->irq); - xtensa_irq_unmask(d->irq); -} + unsigned mask = XCHAL_INTTYPE_MASK_EXTERN_EDGE | + XCHAL_INTTYPE_MASK_EXTERN_LEVEL; + unsigned i; -static void xtensa_irq_disable(struct irq_data *d) -{ - xtensa_irq_mask(d->irq); - variant_irq_disable(d->irq); + for (i = 0; mask; ++i, mask >>= 1) { + if ((mask & 1) && ext_irq-- == 0) + return i; + } + return XCHAL_NUM_INTERRUPTS; } -static void xtensa_irq_ack(struct irq_data *d) +unsigned xtensa_get_ext_irq_no(unsigned irq) { - set_sr(1 << d->irq, INTCLEAR); + unsigned mask = (XCHAL_INTTYPE_MASK_EXTERN_EDGE | + XCHAL_INTTYPE_MASK_EXTERN_LEVEL) & + ((1u << irq) - 1); + return hweight32(mask); } -static int xtensa_irq_retrigger(struct irq_data *d) +void __init init_IRQ(void) { - set_sr (1 << d->irq, INTSET); - return 1; -} - +#ifdef CONFIG_OF + irqchip_init(); +#else +#ifdef CONFIG_HAVE_SMP + xtensa_mx_init_legacy(NULL); +#else + xtensa_pic_init_legacy(NULL); +#endif +#endif -static struct irq_chip xtensa_irq_chip = { - .name = "xtensa", - .irq_enable = xtensa_irq_enable, - .irq_disable = xtensa_irq_disable, - .irq_mask = xtensa_irq_mask, - .irq_unmask = xtensa_irq_unmask, - .irq_ack = xtensa_irq_ack, - .irq_retrigger = xtensa_irq_retrigger, -}; +#ifdef CONFIG_SMP + ipi_init(); +#endif + variant_init_irq(); +} -void __init init_IRQ(void) +#ifdef CONFIG_HOTPLUG_CPU +/* + * The CPU has been marked offline. Migrate IRQs off this CPU. If + * the affinity settings do not allow other CPUs, force them onto any + * available CPU. + */ +void migrate_irqs(void) { - int index; - - for (index = 0; index < XTENSA_NR_IRQS; index++) { - int mask = 1 << index; + unsigned int i, cpu = smp_processor_id(); - if (mask & XCHAL_INTTYPE_MASK_SOFTWARE) - irq_set_chip_and_handler(index, &xtensa_irq_chip, - handle_simple_irq); + for_each_active_irq(i) { + struct irq_data *data = irq_get_irq_data(i); + unsigned int newcpu; - else if (mask & XCHAL_INTTYPE_MASK_EXTERN_EDGE) - irq_set_chip_and_handler(index, &xtensa_irq_chip, - handle_edge_irq); + if (irqd_is_per_cpu(data)) + continue; - else if (mask & XCHAL_INTTYPE_MASK_EXTERN_LEVEL) - irq_set_chip_and_handler(index, &xtensa_irq_chip, - handle_level_irq); + if (!cpumask_test_cpu(cpu, data->affinity)) + continue; - else if (mask & XCHAL_INTTYPE_MASK_TIMER) - irq_set_chip_and_handler(index, &xtensa_irq_chip, - handle_edge_irq); + newcpu = cpumask_any_and(data->affinity, cpu_online_mask); - else /* XCHAL_INTTYPE_MASK_WRITE_ERROR */ - /* XCHAL_INTTYPE_MASK_NMI */ + if (newcpu >= nr_cpu_ids) { + pr_info_ratelimited("IRQ%u no longer affine to CPU%u\n", + i, cpu); - irq_set_chip_and_handler(index, &xtensa_irq_chip, - handle_level_irq); + cpumask_setall(data->affinity); + } + irq_set_affinity(i, data->affinity); } - - cached_irq_mask = 0; - - variant_init_irq(); } +#endif /* CONFIG_HOTPLUG_CPU */ diff --git a/arch/xtensa/kernel/mcount.S b/arch/xtensa/kernel/mcount.S new file mode 100644 index 00000000000..0eeda2e4a25 --- /dev/null +++ b/arch/xtensa/kernel/mcount.S @@ -0,0 +1,50 @@ +/* + * arch/xtensa/kernel/mcount.S + * + * Xtensa specific mcount support + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2013 Tensilica Inc. + */ + +#include <linux/linkage.h> +#include <asm/ftrace.h> + +/* + * Entry condition: + * + * a2: a0 of the caller + */ + +ENTRY(_mcount) + + entry a1, 16 + + movi a4, ftrace_trace_function + l32i a4, a4, 0 + movi a3, ftrace_stub + bne a3, a4, 1f + retw + +1: xor a7, a2, a1 + movi a3, 0x3fffffff + and a7, a7, a3 + xor a7, a7, a1 + + xor a6, a0, a1 + and a6, a6, a3 + xor a6, a6, a1 + addi a6, a6, -MCOUNT_INSN_SIZE + callx4 a4 + + retw + +ENDPROC(_mcount) + +ENTRY(ftrace_stub) + entry a1, 16 + retw +ENDPROC(ftrace_stub) diff --git a/arch/xtensa/kernel/module.c b/arch/xtensa/kernel/module.c index 451dda928c9..b715237bae6 100644 --- a/arch/xtensa/kernel/module.c +++ b/arch/xtensa/kernel/module.c @@ -53,7 +53,7 @@ int apply_relocate_add(Elf32_Shdr *sechdrs, struct module *mod) { unsigned int i; - Elf32_Rela *rela = (void *)sechdrs[relsec].sh_addr; + Elf32_Rela *rela = (void *)sechdrs[relsec].sh_addr; Elf32_Sym *sym; unsigned char *location; uint32_t value; diff --git a/arch/xtensa/kernel/mxhead.S b/arch/xtensa/kernel/mxhead.S new file mode 100644 index 00000000000..77a161a112c --- /dev/null +++ b/arch/xtensa/kernel/mxhead.S @@ -0,0 +1,85 @@ +/* + * Xtensa Secondary Processors startup code. + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2001 - 2013 Tensilica Inc. + * + * Joe Taylor <joe@tensilica.com> + * Chris Zankel <chris@zankel.net> + * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca> + * Pete Delaney <piet@tensilica.com> + */ + +#include <linux/linkage.h> + +#include <asm/cacheasm.h> +#include <asm/initialize_mmu.h> +#include <asm/mxregs.h> +#include <asm/regs.h> + + + .section .SecondaryResetVector.text, "ax" + + +ENTRY(_SecondaryResetVector) + _j _SetupOCD + + .begin no-absolute-literals + .literal_position + +_SetupOCD: + /* + * Initialize WB, WS, and clear PS.EXCM (to allow loop instructions). + * Set Interrupt Level just below XCHAL_DEBUGLEVEL to allow + * xt-gdb to single step via DEBUG exceptions received directly + * by ocd. + */ + movi a1, 1 + movi a0, 0 + wsr a1, windowstart + wsr a0, windowbase + rsync + + movi a1, LOCKLEVEL + wsr a1, ps + rsync + +_SetupMMU: + Offset = _SetupMMU - _SecondaryResetVector + +#ifdef CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX + initialize_mmu +#endif + + /* + * Start Secondary Processors with NULL pointer to boot params. + */ + movi a2, 0 # a2 == NULL + movi a3, _startup + jx a3 + + .end no-absolute-literals + + + .section .SecondaryResetVector.remapped_text, "ax" + .global _RemappedSecondaryResetVector + + .org 0 # Need to do org before literals + +_RemappedSecondaryResetVector: + .begin no-absolute-literals + .literal_position + + _j _RemappedSetupMMU + . = _RemappedSecondaryResetVector + Offset + +_RemappedSetupMMU: + +#ifdef CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX + initialize_mmu +#endif + + .end no-absolute-literals diff --git a/arch/xtensa/kernel/pci-dma.c b/arch/xtensa/kernel/pci-dma.c index 2783fda76dd..2d9cc6dbfd7 100644 --- a/arch/xtensa/kernel/pci-dma.c +++ b/arch/xtensa/kernel/pci-dma.c @@ -21,6 +21,7 @@ #include <linux/string.h> #include <linux/pci.h> #include <linux/gfp.h> +#include <linux/module.h> #include <asm/io.h> #include <asm/cacheflush.h> @@ -62,6 +63,7 @@ dma_alloc_coherent(struct device *dev,size_t size,dma_addr_t *handle,gfp_t flag) return (void*)uncached; } +EXPORT_SYMBOL(dma_alloc_coherent); void dma_free_coherent(struct device *hwdev, size_t size, void *vaddr, dma_addr_t dma_handle) @@ -73,6 +75,7 @@ void dma_free_coherent(struct device *hwdev, size_t size, free_pages(addr, get_order(size)); } +EXPORT_SYMBOL(dma_free_coherent); void consistent_sync(void *vaddr, size_t size, int direction) @@ -92,3 +95,4 @@ void consistent_sync(void *vaddr, size_t size, int direction) break; } } +EXPORT_SYMBOL(consistent_sync); diff --git a/arch/xtensa/kernel/pci.c b/arch/xtensa/kernel/pci.c index 69759e9cb3e..5b3403388d7 100644 --- a/arch/xtensa/kernel/pci.c +++ b/arch/xtensa/kernel/pci.c @@ -77,9 +77,9 @@ pcibios_align_resource(void *data, const struct resource *res, if (res->flags & IORESOURCE_IO) { if (size > 0x100) { - printk(KERN_ERR "PCI: I/O Region %s/%d too large" - " (%ld bytes)\n", pci_name(dev), - dev->resource - res, size); + pr_err("PCI: I/O Region %s/%d too large (%u bytes)\n", + pci_name(dev), dev->resource - res, + size); } if (start & 0x300) @@ -174,7 +174,7 @@ static int __init pcibios_init(void) struct pci_controller *pci_ctrl; struct list_head resources; struct pci_bus *bus; - int next_busno = 0, i; + int next_busno = 0; printk("PCI: Probing PCI hardware\n"); @@ -197,7 +197,7 @@ static int __init pcibios_init(void) subsys_initcall(pcibios_init); -void __init pcibios_fixup_bus(struct pci_bus *bus) +void pcibios_fixup_bus(struct pci_bus *bus) { if (bus->parent) { /* This is a subordinate bridge */ @@ -210,14 +210,6 @@ void pcibios_set_master(struct pci_dev *dev) /* No special bus mastering setup handling */ } -/* the next one is stolen from the alpha port... */ - -void __init -pcibios_update_irq(struct pci_dev *dev, int irq) -{ - pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq); -} - int pcibios_enable_device(struct pci_dev *dev, int mask) { u16 cmd, old_cmd; @@ -341,7 +333,7 @@ __pci_mmap_set_pgprot(struct pci_dev *dev, struct vm_area_struct *vma, int prot = pgprot_val(vma->vm_page_prot); /* Set to write-through */ - prot &= ~_PAGE_NO_CACHE; + prot = (prot & _PAGE_CA_MASK) | _PAGE_CA_WT; #if 0 if (!write_combine) prot |= _PAGE_WRITETHRU; diff --git a/arch/xtensa/kernel/platform.c b/arch/xtensa/kernel/platform.c index 1b91a97f1d8..1cf008284dd 100644 --- a/arch/xtensa/kernel/platform.c +++ b/arch/xtensa/kernel/platform.c @@ -29,19 +29,18 @@ */ _F(void, setup, (char** cmd), { }); -_F(void, init_irq, (void), { }); _F(void, restart, (void), { while(1); }); _F(void, halt, (void), { while(1); }); _F(void, power_off, (void), { while(1); }); _F(void, idle, (void), { __asm__ __volatile__ ("waiti 0" ::: "memory"); }); _F(void, heartbeat, (void), { }); _F(int, pcibios_fixup, (void), { return 0; }); +_F(void, pcibios_init, (void), { }); #ifdef CONFIG_XTENSA_CALIBRATE_CCOUNT _F(void, calibrate_ccount, (void), { - printk ("ERROR: Cannot calibrate cpu frequency! Assuming 100MHz.\n"); - ccount_per_jiffy = 100 * (1000000UL/HZ); + pr_err("ERROR: Cannot calibrate cpu frequency! Assuming 10MHz.\n"); + ccount_freq = 10 * 1000000UL; }); #endif - diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c index 2c8d6a3d250..1c85323f01d 100644 --- a/arch/xtensa/kernel/process.c +++ b/arch/xtensa/kernel/process.c @@ -31,6 +31,7 @@ #include <linux/mqueue.h> #include <linux/fs.h> #include <linux/slab.h> +#include <linux/rcupdate.h> #include <asm/pgtable.h> #include <asm/uaccess.h> @@ -44,6 +45,7 @@ #include <asm/regs.h> extern void ret_from_fork(void); +extern void ret_from_kernel_thread(void); struct task_struct *current_set[NR_CPUS] = {&init_task, }; @@ -103,17 +105,9 @@ void coprocessor_flush_all(struct thread_info *ti) /* * Powermanagement idle function, if any is provided by the platform. */ - -void cpu_idle(void) +void arch_cpu_idle(void) { - local_irq_enable(); - - /* endless idle loop with no priority at all */ - while (1) { - while (!need_resched()) - platform_idle(); - schedule_preempt_disabled(); - } + platform_idle(); } /* @@ -155,64 +149,123 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) /* * Copy thread. * + * There are two modes in which this function is called: + * 1) Userspace thread creation, + * regs != NULL, usp_thread_fn is userspace stack pointer. + * It is expected to copy parent regs (in case CLONE_VM is not set + * in the clone_flags) and set up passed usp in the childregs. + * 2) Kernel thread creation, + * regs == NULL, usp_thread_fn is the function to run in the new thread + * and thread_fn_arg is its parameter. + * childregs are not used for the kernel threads. + * * The stack layout for the new thread looks like this: * - * +------------------------+ <- sp in childregs (= tos) + * +------------------------+ * | childregs | * +------------------------+ <- thread.sp = sp in dummy-frame * | dummy-frame | (saved in dummy-frame spill-area) * +------------------------+ * - * We create a dummy frame to return to ret_from_fork: - * a0 points to ret_from_fork (simulating a call4) + * We create a dummy frame to return to either ret_from_fork or + * ret_from_kernel_thread: + * a0 points to ret_from_fork/ret_from_kernel_thread (simulating a call4) * sp points to itself (thread.sp) - * a2, a3 are unused. + * a2, a3 are unused for userspace threads, + * a2 points to thread_fn, a3 holds thread_fn arg for kernel threads. * * Note: This is a pristine frame, so we don't need any spill region on top of * childregs. + * + * The fun part: if we're keeping the same VM (i.e. cloning a thread, + * not an entire process), we're normally given a new usp, and we CANNOT share + * any live address register windows. If we just copy those live frames over, + * the two threads (parent and child) will overflow the same frames onto the + * parent stack at different times, likely corrupting the parent stack (esp. + * if the parent returns from functions that called clone() and calls new + * ones, before the child overflows its now old copies of its parent windows). + * One solution is to spill windows to the parent stack, but that's fairly + * involved. Much simpler to just not copy those live frames across. */ -int copy_thread(unsigned long clone_flags, unsigned long usp, - unsigned long unused, - struct task_struct * p, struct pt_regs * regs) +int copy_thread(unsigned long clone_flags, unsigned long usp_thread_fn, + unsigned long thread_fn_arg, struct task_struct *p) { - struct pt_regs *childregs; - struct thread_info *ti; - unsigned long tos; - int user_mode = user_mode(regs); + struct pt_regs *childregs = task_pt_regs(p); - /* Set up new TSS. */ - tos = (unsigned long)task_stack_page(p) + THREAD_SIZE; - if (user_mode) - childregs = (struct pt_regs*)(tos - PT_USER_SIZE); - else - childregs = (struct pt_regs*)tos - 1; - - *childregs = *regs; +#if (XTENSA_HAVE_COPROCESSORS || XTENSA_HAVE_IO_PORTS) + struct thread_info *ti; +#endif /* Create a call4 dummy-frame: a0 = 0, a1 = childregs. */ *((int*)childregs - 3) = (unsigned long)childregs; *((int*)childregs - 4) = 0; - childregs->areg[1] = tos; - childregs->areg[2] = 0; - p->set_child_tid = p->clear_child_tid = NULL; - p->thread.ra = MAKE_RA_FOR_CALL((unsigned long)ret_from_fork, 0x1); p->thread.sp = (unsigned long)childregs; - if (user_mode(regs)) { + if (!(p->flags & PF_KTHREAD)) { + struct pt_regs *regs = current_pt_regs(); + unsigned long usp = usp_thread_fn ? + usp_thread_fn : regs->areg[1]; + + p->thread.ra = MAKE_RA_FOR_CALL( + (unsigned long)ret_from_fork, 0x1); - int len = childregs->wmask & ~0xf; + /* This does not copy all the regs. + * In a bout of brilliance or madness, + * ARs beyond a0-a15 exist past the end of the struct. + */ + *childregs = *regs; childregs->areg[1] = usp; - memcpy(&childregs->areg[XCHAL_NUM_AREGS - len/4], - ®s->areg[XCHAL_NUM_AREGS - len/4], len); -// FIXME: we need to set THREADPTR in thread_info... - if (clone_flags & CLONE_SETTLS) - childregs->areg[2] = childregs->areg[6]; + childregs->areg[2] = 0; + + /* When sharing memory with the parent thread, the child + usually starts on a pristine stack, so we have to reset + windowbase, windowstart and wmask. + (Note that such a new thread is required to always create + an initial call4 frame) + The exception is vfork, where the new thread continues to + run on the parent's stack until it calls execve. This could + be a call8 or call12, which requires a legal stack frame + of the previous caller for the overflow handlers to work. + (Note that it's always legal to overflow live registers). + In this case, ensure to spill at least the stack pointer + of that frame. */ + + if (clone_flags & CLONE_VM) { + /* check that caller window is live and same stack */ + int len = childregs->wmask & ~0xf; + if (regs->areg[1] == usp && len != 0) { + int callinc = (regs->areg[0] >> 30) & 3; + int caller_ars = XCHAL_NUM_AREGS - callinc * 4; + put_user(regs->areg[caller_ars+1], + (unsigned __user*)(usp - 12)); + } + childregs->wmask = 1; + childregs->windowstart = 1; + childregs->windowbase = 0; + } else { + int len = childregs->wmask & ~0xf; + memcpy(&childregs->areg[XCHAL_NUM_AREGS - len/4], + ®s->areg[XCHAL_NUM_AREGS - len/4], len); + } + /* The thread pointer is passed in the '4th argument' (= a5) */ + if (clone_flags & CLONE_SETTLS) + childregs->threadptr = childregs->areg[5]; } else { - /* In kernel space, we start a new thread with a new stack. */ - childregs->wmask = 1; + p->thread.ra = MAKE_RA_FOR_CALL( + (unsigned long)ret_from_kernel_thread, 1); + + /* pass parameters to ret_from_kernel_thread: + * a2 = thread_fn, a3 = thread_fn arg + */ + *((int *)childregs - 1) = thread_fn_arg; + *((int *)childregs - 2) = usp_thread_fn; + + /* Childregs are only used when we're going to userspace + * in which case start_thread will set them up. + */ } #if (XTENSA_HAVE_COPROCESSORS || XTENSA_HAVE_IO_PORTS) @@ -301,39 +354,3 @@ int dump_fpu(void) { return 0; } - -asmlinkage -long xtensa_clone(unsigned long clone_flags, unsigned long newsp, - void __user *parent_tid, void *child_tls, - void __user *child_tid, long a5, - struct pt_regs *regs) -{ - if (!newsp) - newsp = regs->areg[1]; - return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); -} - -/* - * xtensa_execve() executes a new program. - */ - -asmlinkage -long xtensa_execve(const char __user *name, - const char __user *const __user *argv, - const char __user *const __user *envp, - long a3, long a4, long a5, - struct pt_regs *regs) -{ - long error; - char * filename; - - filename = getname(name); - error = PTR_ERR(filename); - if (IS_ERR(filename)) - goto out; - error = do_execve(filename, argv, envp, regs); - putname(filename); -out: - return error; -} - diff --git a/arch/xtensa/kernel/ptrace.c b/arch/xtensa/kernel/ptrace.c index 33eea4c16f1..562fac66475 100644 --- a/arch/xtensa/kernel/ptrace.c +++ b/arch/xtensa/kernel/ptrace.c @@ -53,9 +53,8 @@ int ptrace_getregs(struct task_struct *child, void __user *uregs) { struct pt_regs *regs = task_pt_regs(child); xtensa_gregset_t __user *gregset = uregs; - unsigned long wm = regs->wmask; unsigned long wb = regs->windowbase; - int live, i; + int i; if (!access_ok(VERIFY_WRITE, uregs, sizeof(xtensa_gregset_t))) return -EIO; @@ -67,13 +66,11 @@ int ptrace_getregs(struct task_struct *child, void __user *uregs) __put_user(regs->lcount, &gregset->lcount); __put_user(regs->windowstart, &gregset->windowstart); __put_user(regs->windowbase, &gregset->windowbase); + __put_user(regs->threadptr, &gregset->threadptr); - live = (wm & 2) ? 4 : (wm & 4) ? 8 : (wm & 8) ? 12 : 16; - - for (i = 0; i < live; i++) - __put_user(regs->areg[i],gregset->a+((wb*4+i)%XCHAL_NUM_AREGS)); - for (i = XCHAL_NUM_AREGS - (wm >> 4) * 4; i < XCHAL_NUM_AREGS; i++) - __put_user(regs->areg[i],gregset->a+((wb*4+i)%XCHAL_NUM_AREGS)); + for (i = 0; i < XCHAL_NUM_AREGS; i++) + __put_user(regs->areg[i], + gregset->a + ((wb * 4 + i) % XCHAL_NUM_AREGS)); return 0; } @@ -84,7 +81,7 @@ int ptrace_setregs(struct task_struct *child, void __user *uregs) xtensa_gregset_t *gregset = uregs; const unsigned long ps_mask = PS_CALLINC_MASK | PS_OWB_MASK; unsigned long ps; - unsigned long wb; + unsigned long wb, ws; if (!access_ok(VERIFY_WRITE, uregs, sizeof(xtensa_gregset_t))) return -EIO; @@ -94,21 +91,33 @@ int ptrace_setregs(struct task_struct *child, void __user *uregs) __get_user(regs->lbeg, &gregset->lbeg); __get_user(regs->lend, &gregset->lend); __get_user(regs->lcount, &gregset->lcount); - __get_user(regs->windowstart, &gregset->windowstart); + __get_user(ws, &gregset->windowstart); __get_user(wb, &gregset->windowbase); + __get_user(regs->threadptr, &gregset->threadptr); regs->ps = (regs->ps & ~ps_mask) | (ps & ps_mask) | (1 << PS_EXCM_BIT); if (wb >= XCHAL_NUM_AREGS / 4) return -EFAULT; - regs->windowbase = wb; + if (wb != regs->windowbase || ws != regs->windowstart) { + unsigned long rotws, wmask; + + rotws = (((ws | (ws << WSBITS)) >> wb) & + ((1 << WSBITS) - 1)) & ~1; + wmask = ((rotws ? WSBITS + 1 - ffs(rotws) : 0) << 4) | + (rotws & 0xF) | 1; + regs->windowbase = wb; + regs->windowstart = ws; + regs->wmask = wmask; + } if (wb != 0 && __copy_from_user(regs->areg + XCHAL_NUM_AREGS - wb * 4, - gregset->a, wb * 16)) + gregset->a, wb * 16)) return -EFAULT; - if (__copy_from_user(regs->areg, gregset->a + wb*4, (WSBITS-wb) * 16)) + if (__copy_from_user(regs->areg, gregset->a + wb * 4, + (WSBITS - wb) * 16)) return -EFAULT; return 0; @@ -154,7 +163,7 @@ int ptrace_setxregs(struct task_struct *child, void __user *uregs) coprocessor_flush_all(ti); coprocessor_release_all(ti); - ret |= __copy_from_user(&ti->xtregs_cp, &xtregs->cp0, + ret |= __copy_from_user(&ti->xtregs_cp, &xtregs->cp0, sizeof(xtregs_coprocessor_t)); #endif ret |= __copy_from_user(®s->xtregs_opt, &xtregs->opt, @@ -343,4 +352,3 @@ void do_syscall_trace_leave(struct pt_regs *regs) && (current->ptrace & PT_PTRACED)) do_syscall_trace(); } - diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c index 17e746f7be6..06370ccea9e 100644 --- a/arch/xtensa/kernel/setup.c +++ b/arch/xtensa/kernel/setup.c @@ -21,6 +21,11 @@ #include <linux/screen_info.h> #include <linux/bootmem.h> #include <linux/kernel.h> +#include <linux/percpu.h> +#include <linux/clk-provider.h> +#include <linux/cpu.h> +#include <linux/of_fdt.h> +#include <linux/of_platform.h> #if defined(CONFIG_VGA_CONSOLE) || defined(CONFIG_DUMMY_CONSOLE) # include <linux/console.h> @@ -35,6 +40,7 @@ #endif #include <asm/bootparam.h> +#include <asm/mmu_context.h> #include <asm/pgtable.h> #include <asm/processor.h> #include <asm/timex.h> @@ -42,6 +48,9 @@ #include <asm/page.h> #include <asm/setup.h> #include <asm/param.h> +#include <asm/traps.h> +#include <asm/smp.h> +#include <asm/sysmem.h> #include <platform/hardware.h> @@ -58,14 +67,16 @@ extern struct rtc_ops no_rtc_ops; struct rtc_ops *rtc_ops; #ifdef CONFIG_BLK_DEV_INITRD -extern void *initrd_start; -extern void *initrd_end; -extern void *__initrd_start; -extern void *__initrd_end; +extern unsigned long initrd_start; +extern unsigned long initrd_end; int initrd_is_mapped = 0; extern int initrd_below_start_ok; #endif +#ifdef CONFIG_OF +void *dtb_start = __dtb_start; +#endif + unsigned char aux_device_present; extern unsigned long loops_per_jiffy; @@ -77,20 +88,6 @@ static char __initdata command_line[COMMAND_LINE_SIZE]; static char default_command_line[COMMAND_LINE_SIZE] __initdata = CONFIG_CMDLINE; #endif -sysmem_info_t __initdata sysmem; - -#ifdef CONFIG_BLK_DEV_INITRD -int initrd_is_mapped; -#endif - -#ifdef CONFIG_MMU -extern void init_mmu(void); -#else -static inline void init_mmu(void) { } -#endif - -extern void zones_init(void); - /* * Boot parameter parsing. * @@ -106,30 +103,18 @@ typedef struct tagtable { } tagtable_t; #define __tagtable(tag, fn) static tagtable_t __tagtable_##fn \ - __attribute__((unused, __section__(".taglist"))) = { tag, fn } + __attribute__((used, section(".taglist"))) = { tag, fn } /* parse current tag */ static int __init parse_tag_mem(const bp_tag_t *tag) { - meminfo_t *mi = (meminfo_t*)(tag->data); + struct bp_meminfo *mi = (struct bp_meminfo *)(tag->data); if (mi->type != MEMORY_TYPE_CONVENTIONAL) return -1; - if (sysmem.nr_banks >= SYSMEM_BANKS_MAX) { - printk(KERN_WARNING - "Ignoring memory bank 0x%08lx size %ldKB\n", - (unsigned long)mi->start, - (unsigned long)mi->end - (unsigned long)mi->start); - return -EINVAL; - } - sysmem.bank[sysmem.nr_banks].type = mi->type; - sysmem.bank[sysmem.nr_banks].start = PAGE_ALIGN(mi->start); - sysmem.bank[sysmem.nr_banks].end = mi->end & PAGE_SIZE; - sysmem.nr_banks++; - - return 0; + return add_sysmem_bank(mi->start, mi->end); } __tagtable(BP_TAG_MEMORY, parse_tag_mem); @@ -138,22 +123,33 @@ __tagtable(BP_TAG_MEMORY, parse_tag_mem); static int __init parse_tag_initrd(const bp_tag_t* tag) { - meminfo_t* mi; - mi = (meminfo_t*)(tag->data); - initrd_start = (void*)(mi->start); - initrd_end = (void*)(mi->end); + struct bp_meminfo *mi = (struct bp_meminfo *)(tag->data); + + initrd_start = (unsigned long)__va(mi->start); + initrd_end = (unsigned long)__va(mi->end); return 0; } __tagtable(BP_TAG_INITRD, parse_tag_initrd); +#ifdef CONFIG_OF + +static int __init parse_tag_fdt(const bp_tag_t *tag) +{ + dtb_start = __va(tag->data[0]); + return 0; +} + +__tagtable(BP_TAG_FDT, parse_tag_fdt); + +#endif /* CONFIG_OF */ + #endif /* CONFIG_BLK_DEV_INITRD */ static int __init parse_tag_cmdline(const bp_tag_t* tag) { - strncpy(command_line, (char*)(tag->data), COMMAND_LINE_SIZE); - command_line[COMMAND_LINE_SIZE - 1] = '\0'; + strlcpy(command_line, (char *)(tag->data), COMMAND_LINE_SIZE); return 0; } @@ -191,36 +187,108 @@ static int __init parse_bootparam(const bp_tag_t* tag) return 0; } -/* - * Initialize architecture. (Early stage) - */ +#ifdef CONFIG_OF +bool __initdata dt_memory_scan = false; -void __init init_arch(bp_tag_t *bp_start) +#if XCHAL_HAVE_PTP_MMU && XCHAL_HAVE_SPANNING_WAY +unsigned long xtensa_kio_paddr = XCHAL_KIO_DEFAULT_PADDR; +EXPORT_SYMBOL(xtensa_kio_paddr); + +static int __init xtensa_dt_io_area(unsigned long node, const char *uname, + int depth, void *data) { + const __be32 *ranges; + int len; -#ifdef CONFIG_BLK_DEV_INITRD - initrd_start = &__initrd_start; - initrd_end = &__initrd_end; -#endif + if (depth > 1) + return 0; - sysmem.nr_banks = 0; + if (!of_flat_dt_is_compatible(node, "simple-bus")) + return 0; -#ifdef CONFIG_CMDLINE_BOOL - strcpy(command_line, default_command_line); + ranges = of_get_flat_dt_prop(node, "ranges", &len); + if (!ranges) + return 1; + if (len == 0) + return 1; + + xtensa_kio_paddr = of_read_ulong(ranges+1, 1); + /* round down to nearest 256MB boundary */ + xtensa_kio_paddr &= 0xf0000000; + + return 1; +} +#else +static int __init xtensa_dt_io_area(unsigned long node, const char *uname, + int depth, void *data) +{ + return 1; +} #endif +void __init early_init_dt_add_memory_arch(u64 base, u64 size) +{ + if (!dt_memory_scan) + return; + + size &= PAGE_MASK; + add_sysmem_bank(base, base + size); +} + +void * __init early_init_dt_alloc_memory_arch(u64 size, u64 align) +{ + return __alloc_bootmem(size, align, 0); +} + +void __init early_init_devtree(void *params) +{ + if (sysmem.nr_banks == 0) + dt_memory_scan = true; + + early_init_dt_scan(params); + of_scan_flat_dt(xtensa_dt_io_area, NULL); + + if (!command_line[0]) + strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); +} + +static int __init xtensa_device_probe(void) +{ + of_clk_init(NULL); + of_platform_populate(NULL, of_default_bus_match_table, NULL, NULL); + return 0; +} + +device_initcall(xtensa_device_probe); + +#endif /* CONFIG_OF */ + +/* + * Initialize architecture. (Early stage) + */ + +void __init init_arch(bp_tag_t *bp_start) +{ /* Parse boot parameters */ - if (bp_start) - parse_bootparam(bp_start); + if (bp_start) + parse_bootparam(bp_start); + +#ifdef CONFIG_OF + early_init_devtree(dtb_start); +#endif if (sysmem.nr_banks == 0) { - sysmem.nr_banks = 1; - sysmem.bank[0].start = PLATFORM_DEFAULT_MEM_START; - sysmem.bank[0].end = PLATFORM_DEFAULT_MEM_START - + PLATFORM_DEFAULT_MEM_SIZE; + add_sysmem_bank(PLATFORM_DEFAULT_MEM_START, + PLATFORM_DEFAULT_MEM_START + + PLATFORM_DEFAULT_MEM_SIZE); } +#ifdef CONFIG_CMDLINE_BOOL + if (!command_line[0]) + strlcpy(command_line, default_command_line, COMMAND_LINE_SIZE); +#endif + /* Early hook for platforms */ platform_init(bp_start); @@ -246,14 +314,146 @@ extern char _UserExceptionVector_literal_start; extern char _UserExceptionVector_text_end; extern char _DoubleExceptionVector_literal_start; extern char _DoubleExceptionVector_text_end; +#if XCHAL_EXCM_LEVEL >= 2 +extern char _Level2InterruptVector_text_start; +extern char _Level2InterruptVector_text_end; +#endif +#if XCHAL_EXCM_LEVEL >= 3 +extern char _Level3InterruptVector_text_start; +extern char _Level3InterruptVector_text_end; +#endif +#if XCHAL_EXCM_LEVEL >= 4 +extern char _Level4InterruptVector_text_start; +extern char _Level4InterruptVector_text_end; +#endif +#if XCHAL_EXCM_LEVEL >= 5 +extern char _Level5InterruptVector_text_start; +extern char _Level5InterruptVector_text_end; +#endif +#if XCHAL_EXCM_LEVEL >= 6 +extern char _Level6InterruptVector_text_start; +extern char _Level6InterruptVector_text_end; +#endif -void __init setup_arch(char **cmdline_p) + + +#ifdef CONFIG_S32C1I_SELFTEST +#if XCHAL_HAVE_S32C1I + +static int __initdata rcw_word, rcw_probe_pc, rcw_exc; + +/* + * Basic atomic compare-and-swap, that records PC of S32C1I for probing. + * + * If *v == cmp, set *v = set. Return previous *v. + */ +static inline int probed_compare_swap(int *v, int cmp, int set) +{ + int tmp; + + __asm__ __volatile__( + " movi %1, 1f\n" + " s32i %1, %4, 0\n" + " wsr %2, scompare1\n" + "1: s32c1i %0, %3, 0\n" + : "=a" (set), "=&a" (tmp) + : "a" (cmp), "a" (v), "a" (&rcw_probe_pc), "0" (set) + : "memory" + ); + return set; +} + +/* Handle probed exception */ + +static void __init do_probed_exception(struct pt_regs *regs, + unsigned long exccause) +{ + if (regs->pc == rcw_probe_pc) { /* exception on s32c1i ? */ + regs->pc += 3; /* skip the s32c1i instruction */ + rcw_exc = exccause; + } else { + do_unhandled(regs, exccause); + } +} + +/* Simple test of S32C1I (soc bringup assist) */ + +static int __init check_s32c1i(void) +{ + int n, cause1, cause2; + void *handbus, *handdata, *handaddr; /* temporarily saved handlers */ + + rcw_probe_pc = 0; + handbus = trap_set_handler(EXCCAUSE_LOAD_STORE_ERROR, + do_probed_exception); + handdata = trap_set_handler(EXCCAUSE_LOAD_STORE_DATA_ERROR, + do_probed_exception); + handaddr = trap_set_handler(EXCCAUSE_LOAD_STORE_ADDR_ERROR, + do_probed_exception); + + /* First try an S32C1I that does not store: */ + rcw_exc = 0; + rcw_word = 1; + n = probed_compare_swap(&rcw_word, 0, 2); + cause1 = rcw_exc; + + /* took exception? */ + if (cause1 != 0) { + /* unclean exception? */ + if (n != 2 || rcw_word != 1) + panic("S32C1I exception error"); + } else if (rcw_word != 1 || n != 1) { + panic("S32C1I compare error"); + } + + /* Then an S32C1I that stores: */ + rcw_exc = 0; + rcw_word = 0x1234567; + n = probed_compare_swap(&rcw_word, 0x1234567, 0xabcde); + cause2 = rcw_exc; + + if (cause2 != 0) { + /* unclean exception? */ + if (n != 0xabcde || rcw_word != 0x1234567) + panic("S32C1I exception error (b)"); + } else if (rcw_word != 0xabcde || n != 0x1234567) { + panic("S32C1I store error"); + } + + /* Verify consistency of exceptions: */ + if (cause1 || cause2) { + pr_warn("S32C1I took exception %d, %d\n", cause1, cause2); + /* If emulation of S32C1I upon bus error gets implemented, + we can get rid of this panic for single core (not SMP) */ + panic("S32C1I exceptions not currently supported"); + } + if (cause1 != cause2) + panic("inconsistent S32C1I exceptions"); + + trap_set_handler(EXCCAUSE_LOAD_STORE_ERROR, handbus); + trap_set_handler(EXCCAUSE_LOAD_STORE_DATA_ERROR, handdata); + trap_set_handler(EXCCAUSE_LOAD_STORE_ADDR_ERROR, handaddr); + return 0; +} + +#else /* XCHAL_HAVE_S32C1I */ + +/* This condition should not occur with a commercially deployed processor. + Display reminder for early engr test or demo chips / FPGA bitstreams */ +static int __init check_s32c1i(void) { - extern int mem_reserve(unsigned long, unsigned long, int); - extern void bootmem_init(void); + pr_warn("Processor configuration lacks atomic compare-and-swap support!\n"); + return 0; +} - memcpy(boot_command_line, command_line, COMMAND_LINE_SIZE); - boot_command_line[COMMAND_LINE_SIZE-1] = '\0'; +#endif /* XCHAL_HAVE_S32C1I */ +early_initcall(check_s32c1i); +#endif /* CONFIG_S32C1I_SELFTEST */ + + +void __init setup_arch(char **cmdline_p) +{ + strlcpy(boot_command_line, command_line, COMMAND_LINE_SIZE); *cmdline_p = command_line; /* Reserve some memory regions */ @@ -261,9 +461,9 @@ void __init setup_arch(char **cmdline_p) #ifdef CONFIG_BLK_DEV_INITRD if (initrd_start < initrd_end) { initrd_is_mapped = mem_reserve(__pa(initrd_start), - __pa(initrd_end), 0); + __pa(initrd_end), 0) == 0; initrd_below_start_ok = 1; - } else { + } else { initrd_start = 0; } #endif @@ -285,10 +485,37 @@ void __init setup_arch(char **cmdline_p) mem_reserve(__pa(&_DoubleExceptionVector_literal_start), __pa(&_DoubleExceptionVector_text_end), 0); +#if XCHAL_EXCM_LEVEL >= 2 + mem_reserve(__pa(&_Level2InterruptVector_text_start), + __pa(&_Level2InterruptVector_text_end), 0); +#endif +#if XCHAL_EXCM_LEVEL >= 3 + mem_reserve(__pa(&_Level3InterruptVector_text_start), + __pa(&_Level3InterruptVector_text_end), 0); +#endif +#if XCHAL_EXCM_LEVEL >= 4 + mem_reserve(__pa(&_Level4InterruptVector_text_start), + __pa(&_Level4InterruptVector_text_end), 0); +#endif +#if XCHAL_EXCM_LEVEL >= 5 + mem_reserve(__pa(&_Level5InterruptVector_text_start), + __pa(&_Level5InterruptVector_text_end), 0); +#endif +#if XCHAL_EXCM_LEVEL >= 6 + mem_reserve(__pa(&_Level6InterruptVector_text_start), + __pa(&_Level6InterruptVector_text_end), 0); +#endif + + parse_early_param(); bootmem_init(); + unflatten_and_copy_device_tree(); + platform_setup(cmdline_p); +#ifdef CONFIG_SMP + smp_init_cpus(); +#endif paging_init(); zones_init(); @@ -306,6 +533,22 @@ void __init setup_arch(char **cmdline_p) #endif } +static DEFINE_PER_CPU(struct cpu, cpu_data); + +static int __init topology_init(void) +{ + int i; + + for_each_possible_cpu(i) { + struct cpu *cpu = &per_cpu(cpu_data, i); + cpu->hotpluggable = !!i; + register_cpu(cpu, i); + } + + return 0; +} +subsys_initcall(topology_init); + void machine_restart(char * cmd) { platform_restart(); @@ -331,21 +574,27 @@ void machine_power_off(void) static int c_show(struct seq_file *f, void *slot) { + char buf[NR_CPUS * 5]; + + cpulist_scnprintf(buf, sizeof(buf), cpu_online_mask); /* high-level stuff */ - seq_printf(f,"processor\t: 0\n" - "vendor_id\t: Tensilica\n" - "model\t\t: Xtensa " XCHAL_HW_VERSION_NAME "\n" - "core ID\t\t: " XCHAL_CORE_ID "\n" - "build ID\t: 0x%x\n" - "byte order\t: %s\n" - "cpu MHz\t\t: %lu.%02lu\n" - "bogomips\t: %lu.%02lu\n", - XCHAL_BUILD_UNIQUE_ID, - XCHAL_HAVE_BE ? "big" : "little", - CCOUNT_PER_JIFFY/(1000000/HZ), - (CCOUNT_PER_JIFFY/(10000/HZ)) % 100, - loops_per_jiffy/(500000/HZ), - (loops_per_jiffy/(5000/HZ)) % 100); + seq_printf(f, "CPU count\t: %u\n" + "CPU list\t: %s\n" + "vendor_id\t: Tensilica\n" + "model\t\t: Xtensa " XCHAL_HW_VERSION_NAME "\n" + "core ID\t\t: " XCHAL_CORE_ID "\n" + "build ID\t: 0x%x\n" + "byte order\t: %s\n" + "cpu MHz\t\t: %lu.%02lu\n" + "bogomips\t: %lu.%02lu\n", + num_online_cpus(), + buf, + XCHAL_BUILD_UNIQUE_ID, + XCHAL_HAVE_BE ? "big" : "little", + ccount_freq/1000000, + (ccount_freq/10000) % 100, + loops_per_jiffy/(500000/HZ), + (loops_per_jiffy/(5000/HZ)) % 100); seq_printf(f,"flags\t\t: " #if XCHAL_HAVE_NMI @@ -393,6 +642,9 @@ c_show(struct seq_file *f, void *slot) #if XCHAL_HAVE_FP "fpu " #endif +#if XCHAL_HAVE_S32C1I + "s32c1i " +#endif "\n"); /* Registers. */ @@ -424,7 +676,7 @@ c_show(struct seq_file *f, void *slot) "icache size\t: %d\n" "icache flags\t: " #if XCHAL_ICACHE_LINE_LOCKABLE - "lock" + "lock " #endif "\n" "dcache line size: %d\n" @@ -432,10 +684,10 @@ c_show(struct seq_file *f, void *slot) "dcache size\t: %d\n" "dcache flags\t: " #if XCHAL_DCACHE_IS_WRITEBACK - "writeback" + "writeback " #endif #if XCHAL_DCACHE_LINE_LOCKABLE - "lock" + "lock " #endif "\n", XCHAL_ICACHE_LINESIZE, @@ -454,7 +706,7 @@ c_show(struct seq_file *f, void *slot) static void * c_start(struct seq_file *f, loff_t *pos) { - return (void *) ((*pos == 0) ? (void *)1 : NULL); + return (*pos == 0) ? (void *)1 : NULL; } static void * @@ -470,11 +722,10 @@ c_stop(struct seq_file *f, void *v) const struct seq_operations cpuinfo_op = { - start: c_start, - next: c_next, - stop: c_stop, - show: c_show + .start = c_start, + .next = c_next, + .stop = c_stop, + .show = c_show, }; #endif /* CONFIG_PROC_FS */ - diff --git a/arch/xtensa/kernel/signal.c b/arch/xtensa/kernel/signal.c index efe4e854b3c..98b67d5f151 100644 --- a/arch/xtensa/kernel/signal.c +++ b/arch/xtensa/kernel/signal.c @@ -19,7 +19,6 @@ #include <linux/errno.h> #include <linux/ptrace.h> #include <linux/personality.h> -#include <linux/freezer.h> #include <linux/tracehook.h> #include <asm/ucontext.h> @@ -213,7 +212,7 @@ restore_sigcontext(struct pt_regs *regs, struct rt_sigframe __user *frame) if (err) return err; - /* The signal handler may have used coprocessors in which + /* The signal handler may have used coprocessors in which * case they are still enabled. We disable them to force a * reloading of the original task's CP state by the lazy * context-switching mechanisms of CP exception handling. @@ -266,7 +265,7 @@ asmlinkage long xtensa_rt_sigreturn(long a0, long a1, long a2, long a3, ret = regs->areg[2]; - if (do_sigaltstack(&frame->uc.uc_stack, NULL, regs->areg[1]) == -EFAULT) + if (restore_altstack(&frame->uc.uc_stack)) goto badframe; return ret; @@ -338,11 +337,11 @@ static int setup_frame(int sig, struct k_sigaction *ka, siginfo_t *info, struct rt_sigframe *frame; int err = 0; int signal; - unsigned long sp, ra; + unsigned long sp, ra, tp; sp = regs->areg[1]; - if ((ka->sa.sa_flags & SA_ONSTACK) != 0 && ! on_sig_stack(sp)) { + if ((ka->sa.sa_flags & SA_ONSTACK) != 0 && sas_ss_flags(sp) == 0) { sp = current->sas_ss_sp + current->sas_ss_size; } @@ -369,11 +368,7 @@ static int setup_frame(int sig, struct k_sigaction *ka, siginfo_t *info, err |= __put_user(0, &frame->uc.uc_flags); err |= __put_user(0, &frame->uc.uc_link); - err |= __put_user((void *)current->sas_ss_sp, - &frame->uc.uc_stack.ss_sp); - err |= __put_user(sas_ss_flags(regs->areg[1]), - &frame->uc.uc_stack.ss_flags); - err |= __put_user(current->sas_ss_size, &frame->uc.uc_stack.ss_size); + err |= __save_altstack(&frame->uc.uc_stack, regs->areg[1]); err |= setup_sigcontext(frame, regs); err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set)); @@ -396,8 +391,9 @@ static int setup_frame(int sig, struct k_sigaction *ka, siginfo_t *info, * Return context not modified until this point. */ - /* Set up registers for signal handler */ - start_thread(regs, (unsigned long) ka->sa.sa_handler, + /* Set up registers for signal handler; preserve the threadptr */ + tp = regs->threadptr; + start_thread(regs, (unsigned long) ka->sa.sa_handler, (unsigned long) frame); /* Set up a stack frame for a call4 @@ -407,6 +403,7 @@ static int setup_frame(int sig, struct k_sigaction *ka, siginfo_t *info, regs->areg[6] = (unsigned long) signal; regs->areg[7] = (unsigned long) &frame->info; regs->areg[8] = (unsigned long) &frame->uc; + regs->threadptr = tp; /* Set access mode to USER_DS. Nomenclature is outdated, but * functionality is used in uaccess.h @@ -425,16 +422,6 @@ give_sigsegv: return -EFAULT; } -asmlinkage long xtensa_sigaltstack(const stack_t __user *uss, - stack_t __user *uoss, - long a2, long a3, long a4, long a5, - struct pt_regs *regs) -{ - return do_sigaltstack(uss, uoss, regs->areg[1]); -} - - - /* * Note that 'init' is a special process: it doesn't get signals it doesn't * want to handle. Thus you cannot kill init even with a SIGKILL even by @@ -527,9 +514,6 @@ static void do_signal(struct pt_regs *regs) void do_notify_resume(struct pt_regs *regs) { - if (!user_mode(regs)) - return; - if (test_thread_flag(TIF_SIGPENDING)) do_signal(regs); diff --git a/arch/xtensa/kernel/smp.c b/arch/xtensa/kernel/smp.c new file mode 100644 index 00000000000..40b5a3771fb --- /dev/null +++ b/arch/xtensa/kernel/smp.c @@ -0,0 +1,607 @@ +/* + * Xtensa SMP support functions. + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2008 - 2013 Tensilica Inc. + * + * Chris Zankel <chris@zankel.net> + * Joe Taylor <joe@tensilica.com> + * Pete Delaney <piet@tensilica.com + */ + +#include <linux/cpu.h> +#include <linux/cpumask.h> +#include <linux/delay.h> +#include <linux/init.h> +#include <linux/interrupt.h> +#include <linux/irqdomain.h> +#include <linux/irq.h> +#include <linux/kdebug.h> +#include <linux/module.h> +#include <linux/reboot.h> +#include <linux/seq_file.h> +#include <linux/smp.h> +#include <linux/thread_info.h> + +#include <asm/cacheflush.h> +#include <asm/kdebug.h> +#include <asm/mmu_context.h> +#include <asm/mxregs.h> +#include <asm/platform.h> +#include <asm/tlbflush.h> +#include <asm/traps.h> + +#ifdef CONFIG_SMP +# if XCHAL_HAVE_S32C1I == 0 +# error "The S32C1I option is required for SMP." +# endif +#endif + +static void system_invalidate_dcache_range(unsigned long start, + unsigned long size); +static void system_flush_invalidate_dcache_range(unsigned long start, + unsigned long size); + +/* IPI (Inter Process Interrupt) */ + +#define IPI_IRQ 0 + +static irqreturn_t ipi_interrupt(int irq, void *dev_id); +static struct irqaction ipi_irqaction = { + .handler = ipi_interrupt, + .flags = IRQF_PERCPU, + .name = "ipi", +}; + +void ipi_init(void) +{ + unsigned irq = irq_create_mapping(NULL, IPI_IRQ); + setup_irq(irq, &ipi_irqaction); +} + +static inline unsigned int get_core_count(void) +{ + /* Bits 18..21 of SYSCFGID contain the core count minus 1. */ + unsigned int syscfgid = get_er(SYSCFGID); + return ((syscfgid >> 18) & 0xf) + 1; +} + +static inline int get_core_id(void) +{ + /* Bits 0...18 of SYSCFGID contain the core id */ + unsigned int core_id = get_er(SYSCFGID); + return core_id & 0x3fff; +} + +void __init smp_prepare_cpus(unsigned int max_cpus) +{ + unsigned i; + + for (i = 0; i < max_cpus; ++i) + set_cpu_present(i, true); +} + +void __init smp_init_cpus(void) +{ + unsigned i; + unsigned int ncpus = get_core_count(); + unsigned int core_id = get_core_id(); + + pr_info("%s: Core Count = %d\n", __func__, ncpus); + pr_info("%s: Core Id = %d\n", __func__, core_id); + + for (i = 0; i < ncpus; ++i) + set_cpu_possible(i, true); +} + +void __init smp_prepare_boot_cpu(void) +{ + unsigned int cpu = smp_processor_id(); + BUG_ON(cpu != 0); + cpu_asid_cache(cpu) = ASID_USER_FIRST; +} + +void __init smp_cpus_done(unsigned int max_cpus) +{ +} + +static int boot_secondary_processors = 1; /* Set with xt-gdb via .xt-gdb */ +static DECLARE_COMPLETION(cpu_running); + +void secondary_start_kernel(void) +{ + struct mm_struct *mm = &init_mm; + unsigned int cpu = smp_processor_id(); + + init_mmu(); + +#ifdef CONFIG_DEBUG_KERNEL + if (boot_secondary_processors == 0) { + pr_debug("%s: boot_secondary_processors:%d; Hanging cpu:%d\n", + __func__, boot_secondary_processors, cpu); + for (;;) + __asm__ __volatile__ ("waiti " __stringify(LOCKLEVEL)); + } + + pr_debug("%s: boot_secondary_processors:%d; Booting cpu:%d\n", + __func__, boot_secondary_processors, cpu); +#endif + /* Init EXCSAVE1 */ + + secondary_trap_init(); + + /* All kernel threads share the same mm context. */ + + atomic_inc(&mm->mm_users); + atomic_inc(&mm->mm_count); + current->active_mm = mm; + cpumask_set_cpu(cpu, mm_cpumask(mm)); + enter_lazy_tlb(mm, current); + + preempt_disable(); + trace_hardirqs_off(); + + calibrate_delay(); + + notify_cpu_starting(cpu); + + secondary_init_irq(); + local_timer_setup(cpu); + + set_cpu_online(cpu, true); + + local_irq_enable(); + + complete(&cpu_running); + + cpu_startup_entry(CPUHP_ONLINE); +} + +static void mx_cpu_start(void *p) +{ + unsigned cpu = (unsigned)p; + unsigned long run_stall_mask = get_er(MPSCORE); + + set_er(run_stall_mask & ~(1u << cpu), MPSCORE); + pr_debug("%s: cpu: %d, run_stall_mask: %lx ---> %lx\n", + __func__, cpu, run_stall_mask, get_er(MPSCORE)); +} + +static void mx_cpu_stop(void *p) +{ + unsigned cpu = (unsigned)p; + unsigned long run_stall_mask = get_er(MPSCORE); + + set_er(run_stall_mask | (1u << cpu), MPSCORE); + pr_debug("%s: cpu: %d, run_stall_mask: %lx ---> %lx\n", + __func__, cpu, run_stall_mask, get_er(MPSCORE)); +} + +#ifdef CONFIG_HOTPLUG_CPU +unsigned long cpu_start_id __cacheline_aligned; +#endif +unsigned long cpu_start_ccount; + +static int boot_secondary(unsigned int cpu, struct task_struct *ts) +{ + unsigned long timeout = jiffies + msecs_to_jiffies(1000); + unsigned long ccount; + int i; + +#ifdef CONFIG_HOTPLUG_CPU + cpu_start_id = cpu; + system_flush_invalidate_dcache_range( + (unsigned long)&cpu_start_id, sizeof(cpu_start_id)); +#endif + smp_call_function_single(0, mx_cpu_start, (void *)cpu, 1); + + for (i = 0; i < 2; ++i) { + do + ccount = get_ccount(); + while (!ccount); + + cpu_start_ccount = ccount; + + while (time_before(jiffies, timeout)) { + mb(); + if (!cpu_start_ccount) + break; + } + + if (cpu_start_ccount) { + smp_call_function_single(0, mx_cpu_stop, + (void *)cpu, 1); + cpu_start_ccount = 0; + return -EIO; + } + } + return 0; +} + +int __cpu_up(unsigned int cpu, struct task_struct *idle) +{ + int ret = 0; + + if (cpu_asid_cache(cpu) == 0) + cpu_asid_cache(cpu) = ASID_USER_FIRST; + + start_info.stack = (unsigned long)task_pt_regs(idle); + wmb(); + + pr_debug("%s: Calling wakeup_secondary(cpu:%d, idle:%p, sp: %08lx)\n", + __func__, cpu, idle, start_info.stack); + + ret = boot_secondary(cpu, idle); + if (ret == 0) { + wait_for_completion_timeout(&cpu_running, + msecs_to_jiffies(1000)); + if (!cpu_online(cpu)) + ret = -EIO; + } + + if (ret) + pr_err("CPU %u failed to boot\n", cpu); + + return ret; +} + +#ifdef CONFIG_HOTPLUG_CPU + +/* + * __cpu_disable runs on the processor to be shutdown. + */ +int __cpu_disable(void) +{ + unsigned int cpu = smp_processor_id(); + + /* + * Take this CPU offline. Once we clear this, we can't return, + * and we must not schedule until we're ready to give up the cpu. + */ + set_cpu_online(cpu, false); + + /* + * OK - migrate IRQs away from this CPU + */ + migrate_irqs(); + + /* + * Flush user cache and TLB mappings, and then remove this CPU + * from the vm mask set of all processes. + */ + local_flush_cache_all(); + local_flush_tlb_all(); + invalidate_page_directory(); + + clear_tasks_mm_cpumask(cpu); + + return 0; +} + +static void platform_cpu_kill(unsigned int cpu) +{ + smp_call_function_single(0, mx_cpu_stop, (void *)cpu, true); +} + +/* + * called on the thread which is asking for a CPU to be shutdown - + * waits until shutdown has completed, or it is timed out. + */ +void __cpu_die(unsigned int cpu) +{ + unsigned long timeout = jiffies + msecs_to_jiffies(1000); + while (time_before(jiffies, timeout)) { + system_invalidate_dcache_range((unsigned long)&cpu_start_id, + sizeof(cpu_start_id)); + if (cpu_start_id == -cpu) { + platform_cpu_kill(cpu); + return; + } + } + pr_err("CPU%u: unable to kill\n", cpu); +} + +void arch_cpu_idle_dead(void) +{ + cpu_die(); +} +/* + * Called from the idle thread for the CPU which has been shutdown. + * + * Note that we disable IRQs here, but do not re-enable them + * before returning to the caller. This is also the behaviour + * of the other hotplug-cpu capable cores, so presumably coming + * out of idle fixes this. + */ +void __ref cpu_die(void) +{ + idle_task_exit(); + local_irq_disable(); + __asm__ __volatile__( + " movi a2, cpu_restart\n" + " jx a2\n"); +} + +#endif /* CONFIG_HOTPLUG_CPU */ + +enum ipi_msg_type { + IPI_RESCHEDULE = 0, + IPI_CALL_FUNC, + IPI_CPU_STOP, + IPI_MAX +}; + +static const struct { + const char *short_text; + const char *long_text; +} ipi_text[] = { + { .short_text = "RES", .long_text = "Rescheduling interrupts" }, + { .short_text = "CAL", .long_text = "Function call interrupts" }, + { .short_text = "DIE", .long_text = "CPU shutdown interrupts" }, +}; + +struct ipi_data { + unsigned long ipi_count[IPI_MAX]; +}; + +static DEFINE_PER_CPU(struct ipi_data, ipi_data); + +static void send_ipi_message(const struct cpumask *callmask, + enum ipi_msg_type msg_id) +{ + int index; + unsigned long mask = 0; + + for_each_cpu(index, callmask) + if (index != smp_processor_id()) + mask |= 1 << index; + + set_er(mask, MIPISET(msg_id)); +} + +void arch_send_call_function_ipi_mask(const struct cpumask *mask) +{ + send_ipi_message(mask, IPI_CALL_FUNC); +} + +void arch_send_call_function_single_ipi(int cpu) +{ + send_ipi_message(cpumask_of(cpu), IPI_CALL_FUNC); +} + +void smp_send_reschedule(int cpu) +{ + send_ipi_message(cpumask_of(cpu), IPI_RESCHEDULE); +} + +void smp_send_stop(void) +{ + struct cpumask targets; + + cpumask_copy(&targets, cpu_online_mask); + cpumask_clear_cpu(smp_processor_id(), &targets); + send_ipi_message(&targets, IPI_CPU_STOP); +} + +static void ipi_cpu_stop(unsigned int cpu) +{ + set_cpu_online(cpu, false); + machine_halt(); +} + +irqreturn_t ipi_interrupt(int irq, void *dev_id) +{ + unsigned int cpu = smp_processor_id(); + struct ipi_data *ipi = &per_cpu(ipi_data, cpu); + unsigned int msg; + unsigned i; + + msg = get_er(MIPICAUSE(cpu)); + for (i = 0; i < IPI_MAX; i++) + if (msg & (1 << i)) { + set_er(1 << i, MIPICAUSE(cpu)); + ++ipi->ipi_count[i]; + } + + if (msg & (1 << IPI_RESCHEDULE)) + scheduler_ipi(); + if (msg & (1 << IPI_CALL_FUNC)) + generic_smp_call_function_interrupt(); + if (msg & (1 << IPI_CPU_STOP)) + ipi_cpu_stop(cpu); + + return IRQ_HANDLED; +} + +void show_ipi_list(struct seq_file *p, int prec) +{ + unsigned int cpu; + unsigned i; + + for (i = 0; i < IPI_MAX; ++i) { + seq_printf(p, "%*s:", prec, ipi_text[i].short_text); + for_each_online_cpu(cpu) + seq_printf(p, " %10lu", + per_cpu(ipi_data, cpu).ipi_count[i]); + seq_printf(p, " %s\n", ipi_text[i].long_text); + } +} + +int setup_profiling_timer(unsigned int multiplier) +{ + pr_debug("setup_profiling_timer %d\n", multiplier); + return 0; +} + +/* TLB flush functions */ + +struct flush_data { + struct vm_area_struct *vma; + unsigned long addr1; + unsigned long addr2; +}; + +static void ipi_flush_tlb_all(void *arg) +{ + local_flush_tlb_all(); +} + +void flush_tlb_all(void) +{ + on_each_cpu(ipi_flush_tlb_all, NULL, 1); +} + +static void ipi_flush_tlb_mm(void *arg) +{ + local_flush_tlb_mm(arg); +} + +void flush_tlb_mm(struct mm_struct *mm) +{ + on_each_cpu(ipi_flush_tlb_mm, mm, 1); +} + +static void ipi_flush_tlb_page(void *arg) +{ + struct flush_data *fd = arg; + local_flush_tlb_page(fd->vma, fd->addr1); +} + +void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr) +{ + struct flush_data fd = { + .vma = vma, + .addr1 = addr, + }; + on_each_cpu(ipi_flush_tlb_page, &fd, 1); +} + +static void ipi_flush_tlb_range(void *arg) +{ + struct flush_data *fd = arg; + local_flush_tlb_range(fd->vma, fd->addr1, fd->addr2); +} + +void flush_tlb_range(struct vm_area_struct *vma, + unsigned long start, unsigned long end) +{ + struct flush_data fd = { + .vma = vma, + .addr1 = start, + .addr2 = end, + }; + on_each_cpu(ipi_flush_tlb_range, &fd, 1); +} + +static void ipi_flush_tlb_kernel_range(void *arg) +{ + struct flush_data *fd = arg; + local_flush_tlb_kernel_range(fd->addr1, fd->addr2); +} + +void flush_tlb_kernel_range(unsigned long start, unsigned long end) +{ + struct flush_data fd = { + .addr1 = start, + .addr2 = end, + }; + on_each_cpu(ipi_flush_tlb_kernel_range, &fd, 1); +} + +/* Cache flush functions */ + +static void ipi_flush_cache_all(void *arg) +{ + local_flush_cache_all(); +} + +void flush_cache_all(void) +{ + on_each_cpu(ipi_flush_cache_all, NULL, 1); +} + +static void ipi_flush_cache_page(void *arg) +{ + struct flush_data *fd = arg; + local_flush_cache_page(fd->vma, fd->addr1, fd->addr2); +} + +void flush_cache_page(struct vm_area_struct *vma, + unsigned long address, unsigned long pfn) +{ + struct flush_data fd = { + .vma = vma, + .addr1 = address, + .addr2 = pfn, + }; + on_each_cpu(ipi_flush_cache_page, &fd, 1); +} + +static void ipi_flush_cache_range(void *arg) +{ + struct flush_data *fd = arg; + local_flush_cache_range(fd->vma, fd->addr1, fd->addr2); +} + +void flush_cache_range(struct vm_area_struct *vma, + unsigned long start, unsigned long end) +{ + struct flush_data fd = { + .vma = vma, + .addr1 = start, + .addr2 = end, + }; + on_each_cpu(ipi_flush_cache_range, &fd, 1); +} + +static void ipi_flush_icache_range(void *arg) +{ + struct flush_data *fd = arg; + local_flush_icache_range(fd->addr1, fd->addr2); +} + +void flush_icache_range(unsigned long start, unsigned long end) +{ + struct flush_data fd = { + .addr1 = start, + .addr2 = end, + }; + on_each_cpu(ipi_flush_icache_range, &fd, 1); +} + +/* ------------------------------------------------------------------------- */ + +static void ipi_invalidate_dcache_range(void *arg) +{ + struct flush_data *fd = arg; + __invalidate_dcache_range(fd->addr1, fd->addr2); +} + +static void system_invalidate_dcache_range(unsigned long start, + unsigned long size) +{ + struct flush_data fd = { + .addr1 = start, + .addr2 = size, + }; + on_each_cpu(ipi_invalidate_dcache_range, &fd, 1); +} + +static void ipi_flush_invalidate_dcache_range(void *arg) +{ + struct flush_data *fd = arg; + __flush_invalidate_dcache_range(fd->addr1, fd->addr2); +} + +static void system_flush_invalidate_dcache_range(unsigned long start, + unsigned long size) +{ + struct flush_data fd = { + .addr1 = start, + .addr2 = size, + }; + on_each_cpu(ipi_flush_invalidate_dcache_range, &fd, 1); +} diff --git a/arch/xtensa/kernel/stacktrace.c b/arch/xtensa/kernel/stacktrace.c new file mode 100644 index 00000000000..7d2c317bd98 --- /dev/null +++ b/arch/xtensa/kernel/stacktrace.c @@ -0,0 +1,120 @@ +/* + * arch/xtensa/kernel/stacktrace.c + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2001 - 2013 Tensilica Inc. + */ +#include <linux/export.h> +#include <linux/sched.h> +#include <linux/stacktrace.h> + +#include <asm/stacktrace.h> +#include <asm/traps.h> + +void walk_stackframe(unsigned long *sp, + int (*fn)(struct stackframe *frame, void *data), + void *data) +{ + unsigned long a0, a1; + unsigned long sp_end; + + a1 = (unsigned long)sp; + sp_end = ALIGN(a1, THREAD_SIZE); + + spill_registers(); + + while (a1 < sp_end) { + struct stackframe frame; + + sp = (unsigned long *)a1; + + a0 = *(sp - 4); + a1 = *(sp - 3); + + if (a1 <= (unsigned long)sp) + break; + + frame.pc = MAKE_PC_FROM_RA(a0, a1); + frame.sp = a1; + + if (fn(&frame, data)) + return; + } +} + +#ifdef CONFIG_STACKTRACE + +struct stack_trace_data { + struct stack_trace *trace; + unsigned skip; +}; + +static int stack_trace_cb(struct stackframe *frame, void *data) +{ + struct stack_trace_data *trace_data = data; + struct stack_trace *trace = trace_data->trace; + + if (trace_data->skip) { + --trace_data->skip; + return 0; + } + if (!kernel_text_address(frame->pc)) + return 0; + + trace->entries[trace->nr_entries++] = frame->pc; + return trace->nr_entries >= trace->max_entries; +} + +void save_stack_trace_tsk(struct task_struct *task, struct stack_trace *trace) +{ + struct stack_trace_data trace_data = { + .trace = trace, + .skip = trace->skip, + }; + walk_stackframe(stack_pointer(task), stack_trace_cb, &trace_data); +} +EXPORT_SYMBOL_GPL(save_stack_trace_tsk); + +void save_stack_trace(struct stack_trace *trace) +{ + save_stack_trace_tsk(current, trace); +} +EXPORT_SYMBOL_GPL(save_stack_trace); + +#endif + +#ifdef CONFIG_FRAME_POINTER + +struct return_addr_data { + unsigned long addr; + unsigned skip; +}; + +static int return_address_cb(struct stackframe *frame, void *data) +{ + struct return_addr_data *r = data; + + if (r->skip) { + --r->skip; + return 0; + } + if (!kernel_text_address(frame->pc)) + return 0; + r->addr = frame->pc; + return 1; +} + +unsigned long return_address(unsigned level) +{ + struct return_addr_data r = { + .skip = level + 1, + }; + walk_stackframe(stack_pointer(NULL), return_address_cb, &r); + return r.addr; +} +EXPORT_SYMBOL(return_address); + +#endif diff --git a/arch/xtensa/kernel/syscall.c b/arch/xtensa/kernel/syscall.c index 05b3f093d5d..5d3f7a119ed 100644 --- a/arch/xtensa/kernel/syscall.c +++ b/arch/xtensa/kernel/syscall.c @@ -32,13 +32,14 @@ typedef void (*syscall_t)(void); syscall_t sys_call_table[__NR_syscall_count] /* FIXME __cacheline_aligned */= { [0 ... __NR_syscall_count - 1] = (syscall_t)&sys_ni_syscall, -#undef __SYSCALL #define __SYSCALL(nr,symbol,nargs) [ nr ] = (syscall_t)symbol, -#undef _XTENSA_UNISTD_H -#undef __KERNEL_SYSCALLS__ -#include <asm/unistd.h> +#include <uapi/asm/unistd.h> }; +#define COLOUR_ALIGN(addr, pgoff) \ + ((((addr) + SHMLBA - 1) & ~(SHMLBA - 1)) + \ + (((pgoff) << PAGE_SHIFT) & (SHMLBA - 1))) + asmlinkage long xtensa_shmat(int shmid, char __user *shmaddr, int shmflg) { unsigned long ret; @@ -50,8 +51,45 @@ asmlinkage long xtensa_shmat(int shmid, char __user *shmaddr, int shmflg) return (long)ret; } -asmlinkage long xtensa_fadvise64_64(int fd, int advice, unsigned long long offset, unsigned long long len) +asmlinkage long xtensa_fadvise64_64(int fd, int advice, + unsigned long long offset, unsigned long long len) { return sys_fadvise64_64(fd, offset, len, advice); } +unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr, + unsigned long len, unsigned long pgoff, unsigned long flags) +{ + struct vm_area_struct *vmm; + + if (flags & MAP_FIXED) { + /* We do not accept a shared mapping if it would violate + * cache aliasing constraints. + */ + if ((flags & MAP_SHARED) && + ((addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1))) + return -EINVAL; + return addr; + } + + if (len > TASK_SIZE) + return -ENOMEM; + if (!addr) + addr = TASK_UNMAPPED_BASE; + + if (flags & MAP_SHARED) + addr = COLOUR_ALIGN(addr, pgoff); + else + addr = PAGE_ALIGN(addr); + + for (vmm = find_vma(current->mm, addr); ; vmm = vmm->vm_next) { + /* At this point: (!vmm || addr < vmm->vm_end). */ + if (TASK_SIZE - len < addr) + return -ENOMEM; + if (!vmm || addr + len <= vmm->vm_start) + return addr; + addr = vmm->vm_end; + if (flags & MAP_SHARED) + addr = COLOUR_ALIGN(addr, pgoff); + } +} diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c index ac62f9cf1e1..2a1823de69c 100644 --- a/arch/xtensa/kernel/time.c +++ b/arch/xtensa/kernel/time.c @@ -16,99 +16,163 @@ #include <linux/sched.h> #include <linux/time.h> #include <linux/clocksource.h> +#include <linux/clockchips.h> #include <linux/interrupt.h> #include <linux/module.h> #include <linux/init.h> #include <linux/irq.h> #include <linux/profile.h> #include <linux/delay.h> +#include <linux/irqdomain.h> +#include <linux/sched_clock.h> #include <asm/timex.h> #include <asm/platform.h> -#ifdef CONFIG_XTENSA_CALIBRATE_CCOUNT -unsigned long ccount_per_jiffy; /* per 1/HZ */ -unsigned long nsec_per_ccount; /* nsec per ccount increment */ -#endif +unsigned long ccount_freq; /* ccount Hz */ +EXPORT_SYMBOL(ccount_freq); -static cycle_t ccount_read(void) +static cycle_t ccount_read(struct clocksource *cs) { return (cycle_t)get_ccount(); } +static u64 notrace ccount_sched_clock_read(void) +{ + return get_ccount(); +} + static struct clocksource ccount_clocksource = { .name = "ccount", .rating = 200, .read = ccount_read, .mask = CLOCKSOURCE_MASK(32), + .flags = CLOCK_SOURCE_IS_CONTINUOUS, +}; + +static int ccount_timer_set_next_event(unsigned long delta, + struct clock_event_device *dev); +static void ccount_timer_set_mode(enum clock_event_mode mode, + struct clock_event_device *evt); +struct ccount_timer { + struct clock_event_device evt; + int irq_enabled; + char name[24]; }; +static DEFINE_PER_CPU(struct ccount_timer, ccount_timer); + +static int ccount_timer_set_next_event(unsigned long delta, + struct clock_event_device *dev) +{ + unsigned long flags, next; + int ret = 0; + + local_irq_save(flags); + next = get_ccount() + delta; + set_linux_timer(next); + if (next - get_ccount() > delta) + ret = -ETIME; + local_irq_restore(flags); + + return ret; +} + +static void ccount_timer_set_mode(enum clock_event_mode mode, + struct clock_event_device *evt) +{ + struct ccount_timer *timer = + container_of(evt, struct ccount_timer, evt); + + /* + * There is no way to disable the timer interrupt at the device level, + * only at the intenable register itself. Since enable_irq/disable_irq + * calls are nested, we need to make sure that these calls are + * balanced. + */ + switch (mode) { + case CLOCK_EVT_MODE_SHUTDOWN: + case CLOCK_EVT_MODE_UNUSED: + if (timer->irq_enabled) { + disable_irq(evt->irq); + timer->irq_enabled = 0; + } + break; + case CLOCK_EVT_MODE_RESUME: + case CLOCK_EVT_MODE_ONESHOT: + if (!timer->irq_enabled) { + enable_irq(evt->irq); + timer->irq_enabled = 1; + } + default: + break; + } +} static irqreturn_t timer_interrupt(int irq, void *dev_id); static struct irqaction timer_irqaction = { .handler = timer_interrupt, - .flags = IRQF_DISABLED, + .flags = IRQF_TIMER, .name = "timer", }; +void local_timer_setup(unsigned cpu) +{ + struct ccount_timer *timer = &per_cpu(ccount_timer, cpu); + struct clock_event_device *clockevent = &timer->evt; + + timer->irq_enabled = 1; + clockevent->name = timer->name; + snprintf(timer->name, sizeof(timer->name), "ccount_clockevent_%u", cpu); + clockevent->features = CLOCK_EVT_FEAT_ONESHOT; + clockevent->rating = 300; + clockevent->set_next_event = ccount_timer_set_next_event; + clockevent->set_mode = ccount_timer_set_mode; + clockevent->cpumask = cpumask_of(cpu); + clockevent->irq = irq_create_mapping(NULL, LINUX_TIMER_INT); + if (WARN(!clockevent->irq, "error: can't map timer irq")) + return; + clockevents_config_and_register(clockevent, ccount_freq, + 0xf, 0xffffffff); +} + void __init time_init(void) { #ifdef CONFIG_XTENSA_CALIBRATE_CCOUNT printk("Calibrating CPU frequency "); platform_calibrate_ccount(); - printk("%d.%02d MHz\n", (int)ccount_per_jiffy/(1000000/HZ), - (int)(ccount_per_jiffy/(10000/HZ))%100); + printk("%d.%02d MHz\n", (int)ccount_freq/1000000, + (int)(ccount_freq/10000)%100); +#else + ccount_freq = CONFIG_XTENSA_CPU_CLOCK*1000000UL; #endif - clocksource_register_hz(&ccount_clocksource, CCOUNT_PER_JIFFY * HZ); - - /* Initialize the linux timer interrupt. */ - - setup_irq(LINUX_TIMER_INT, &timer_irqaction); - set_linux_timer(get_ccount() + CCOUNT_PER_JIFFY); + clocksource_register_hz(&ccount_clocksource, ccount_freq); + local_timer_setup(0); + setup_irq(this_cpu_ptr(&ccount_timer)->evt.irq, &timer_irqaction); + sched_clock_register(ccount_sched_clock_read, 32, ccount_freq); + clocksource_of_init(); } /* * The timer interrupt is called HZ times per second. */ -irqreturn_t timer_interrupt (int irq, void *dev_id) +irqreturn_t timer_interrupt(int irq, void *dev_id) { + struct clock_event_device *evt = &this_cpu_ptr(&ccount_timer)->evt; - unsigned long next; - - next = get_linux_timer(); - -again: - while ((signed long)(get_ccount() - next) > 0) { - - profile_tick(CPU_PROFILING); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif - - xtime_update(1); /* Linux handler in kernel/time/timekeeping */ - - /* Note that writing CCOMPARE clears the interrupt. */ - - next += CCOUNT_PER_JIFFY; - set_linux_timer(next); - } + set_linux_timer(get_linux_timer()); + evt->event_handler(evt); /* Allow platform to do something useful (Wdog). */ - platform_heartbeat(); - /* Make sure we didn't miss any tick... */ - - if ((signed long)(get_ccount() - next) > 0) - goto again; - return IRQ_HANDLED; } #ifndef CONFIG_GENERIC_CALIBRATE_DELAY -void __cpuinit calibrate_delay(void) +void calibrate_delay(void) { - loops_per_jiffy = CCOUNT_PER_JIFFY; + loops_per_jiffy = ccount_freq / HZ; printk("Calibrating delay loop (skipped)... " "%lu.%02lu BogoMIPS preset\n", loops_per_jiffy/(1000000/HZ), diff --git a/arch/xtensa/kernel/traps.c b/arch/xtensa/kernel/traps.c index bc1e14cf936..eebbfd8c26f 100644 --- a/arch/xtensa/kernel/traps.c +++ b/arch/xtensa/kernel/traps.c @@ -11,7 +11,7 @@ * * Essentially rewritten for the Xtensa architecture port. * - * Copyright (C) 2001 - 2005 Tensilica Inc. + * Copyright (C) 2001 - 2013 Tensilica Inc. * * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com> * Chris Zankel <chris@zankel.net> @@ -32,11 +32,13 @@ #include <linux/delay.h> #include <linux/hardirq.h> +#include <asm/stacktrace.h> #include <asm/ptrace.h> #include <asm/timex.h> #include <asm/uaccess.h> #include <asm/pgtable.h> #include <asm/processor.h> +#include <asm/traps.h> #ifdef CONFIG_KGDB extern int gdb_enter; @@ -97,7 +99,7 @@ static dispatch_init_table_t __initdata dispatch_init_table[] = { /* EXCCAUSE_INTEGER_DIVIDE_BY_ZERO unhandled */ /* EXCCAUSE_PRIVILEGED unhandled */ #if XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION -#ifdef CONFIG_UNALIGNED_USER +#ifdef CONFIG_XTENSA_UNALIGNED_USER { EXCCAUSE_UNALIGNED, USER, fast_unaligned }, #else { EXCCAUSE_UNALIGNED, 0, do_unaligned_user }, @@ -155,7 +157,7 @@ COPROCESSOR(7), * 2. it is a temporary memory buffer for the exception handlers. */ -unsigned long exc_table[EXC_TABLE_SIZE/4]; +DEFINE_PER_CPU(unsigned long, exc_table[EXC_TABLE_SIZE/4]); void die(const char*, struct pt_regs*, long); @@ -193,30 +195,48 @@ void do_multihit(struct pt_regs *regs, unsigned long exccause) } /* - * Level-1 interrupt. - * We currently have no priority encoding. + * IRQ handler. */ -unsigned long ignored_level1_interrupts; extern void do_IRQ(int, struct pt_regs *); -void do_interrupt (struct pt_regs *regs) +void do_interrupt(struct pt_regs *regs) { - unsigned long intread = get_sr (INTREAD); - unsigned long intenable = get_sr (INTENABLE); - int i, mask; + static const unsigned int_level_mask[] = { + 0, + XCHAL_INTLEVEL1_MASK, + XCHAL_INTLEVEL2_MASK, + XCHAL_INTLEVEL3_MASK, + XCHAL_INTLEVEL4_MASK, + XCHAL_INTLEVEL5_MASK, + XCHAL_INTLEVEL6_MASK, + XCHAL_INTLEVEL7_MASK, + }; + struct pt_regs *old_regs = set_irq_regs(regs); + + irq_enter(); + + for (;;) { + unsigned intread = get_sr(interrupt); + unsigned intenable = get_sr(intenable); + unsigned int_at_level = intread & intenable; + unsigned level; + + for (level = LOCKLEVEL; level > 0; --level) { + if (int_at_level & int_level_mask[level]) { + int_at_level &= int_level_mask[level]; + break; + } + } - /* Handle all interrupts (no priorities). - * (Clear the interrupt before processing, in case it's - * edge-triggered or software-generated) - */ + if (level == 0) + break; - for (i=0, mask = 1; i < XCHAL_NUM_INTERRUPTS; i++, mask <<= 1) { - if (mask & (intread & intenable)) { - set_sr (mask, INTCLEAR); - do_IRQ (i,regs); - } + do_IRQ(__ffs(int_at_level), regs); } + + irq_exit(); + set_irq_regs(old_regs); } /* @@ -244,7 +264,7 @@ do_illegal_instruction(struct pt_regs *regs) */ #if XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION -#ifndef CONFIG_UNALIGNED_USER +#ifndef CONFIG_XTENSA_UNALIGNED_USER void do_unaligned_user (struct pt_regs *regs) { @@ -293,6 +313,31 @@ do_debug(struct pt_regs *regs) } +static void set_handler(int idx, void *handler) +{ + unsigned int cpu; + + for_each_possible_cpu(cpu) + per_cpu(exc_table, cpu)[idx] = (unsigned long)handler; +} + +/* Set exception C handler - for temporary use when probing exceptions */ + +void * __init trap_set_handler(int cause, void *handler) +{ + void *previous = (void *)per_cpu(exc_table, 0)[ + EXC_TABLE_DEFAULT / 4 + cause]; + set_handler(EXC_TABLE_DEFAULT / 4 + cause, handler); + return previous; +} + + +static void trap_init_excsave(void) +{ + unsigned long excsave1 = (unsigned long)this_cpu_ptr(exc_table); + __asm__ __volatile__("wsr %0, excsave1\n" : : "a" (excsave1)); +} + /* * Initialize dispatch tables. * @@ -306,8 +351,6 @@ do_debug(struct pt_regs *regs) * See vectors.S for more details. */ -#define set_handler(idx,handler) (exc_table[idx] = (unsigned long) (handler)) - void __init trap_init(void) { int i; @@ -337,10 +380,15 @@ void __init trap_init(void) } /* Initialize EXCSAVE_1 to hold the address of the exception table. */ + trap_init_excsave(); +} - i = (unsigned long)exc_table; - __asm__ __volatile__("wsr %0, "__stringify(EXCSAVE_1)"\n" : : "a" (i)); +#ifdef CONFIG_SMP +void secondary_trap_init(void) +{ + trap_init_excsave(); } +#endif /* * This function dumps the current valid window frame and other base registers. @@ -350,6 +398,8 @@ void show_regs(struct pt_regs * regs) { int i, wmask; + show_regs_print_info(KERN_DEFAULT); + wmask = regs->wmask & ~1; for (i = 0; i < 16; i++) { @@ -369,72 +419,25 @@ void show_regs(struct pt_regs * regs) regs->syscall); } -static __always_inline unsigned long *stack_pointer(struct task_struct *task) +static int show_trace_cb(struct stackframe *frame, void *data) { - unsigned long *sp; - - if (!task || task == current) - __asm__ __volatile__ ("mov %0, a1\n" : "=a"(sp)); - else - sp = (unsigned long *)task->thread.sp; - - return sp; -} - -static inline void spill_registers(void) -{ - unsigned int a0, ps; - - __asm__ __volatile__ ( - "movi a14," __stringify (PS_EXCM_BIT) " | 1\n\t" - "mov a12, a0\n\t" - "rsr a13," __stringify(SAR) "\n\t" - "xsr a14," __stringify(PS) "\n\t" - "movi a0, _spill_registers\n\t" - "rsync\n\t" - "callx0 a0\n\t" - "mov a0, a12\n\t" - "wsr a13," __stringify(SAR) "\n\t" - "wsr a14," __stringify(PS) "\n\t" - :: "a" (&a0), "a" (&ps) - : "a2", "a3", "a4", "a7", "a11", "a12", "a13", "a14", "a15", "memory"); + if (kernel_text_address(frame->pc)) { + printk(" [<%08lx>] ", frame->pc); + print_symbol("%s\n", frame->pc); + } + return 0; } void show_trace(struct task_struct *task, unsigned long *sp) { - unsigned long a0, a1, pc; - unsigned long sp_start, sp_end; - - if (sp) - a1 = (unsigned long)sp; - else - a1 = (unsigned long)stack_pointer(task); - - sp_start = a1 & ~(THREAD_SIZE-1); - sp_end = sp_start + THREAD_SIZE; + if (!sp) + sp = stack_pointer(task); printk("Call Trace:"); #ifdef CONFIG_KALLSYMS printk("\n"); #endif - spill_registers(); - - while (a1 > sp_start && a1 < sp_end) { - sp = (unsigned long*)a1; - - a0 = *(sp - 4); - a1 = *(sp - 3); - - if (a1 <= (unsigned long) sp) - break; - - pc = MAKE_PC_FROM_RA(a0, a1); - - if (kernel_text_address(pc)) { - printk(" [<%08lx>] ", pc); - print_symbol("%s\n", pc); - } - } + walk_stackframe(sp, show_trace_cb, NULL); printk("\n"); } @@ -452,7 +455,7 @@ void show_stack(struct task_struct *task, unsigned long *sp) if (!sp) sp = stack_pointer(task); - stack = sp; + stack = sp; printk("\nStack: "); @@ -467,14 +470,6 @@ void show_stack(struct task_struct *task, unsigned long *sp) show_trace(task, stack); } -void dump_stack(void) -{ - show_stack(current, NULL); -} - -EXPORT_SYMBOL(dump_stack); - - void show_code(unsigned int *pc) { long i; @@ -512,7 +507,7 @@ void die(const char * str, struct pt_regs * regs, long err) if (!user_mode(regs)) show_stack(NULL, (unsigned long*)regs->areg[1]); - add_taint(TAINT_DIE); + add_taint(TAINT_DIE, LOCKDEP_NOW_UNRELIABLE); spin_unlock_irq(&die_lock); if (in_interrupt()) @@ -523,5 +518,3 @@ void die(const char * str, struct pt_regs * regs, long err) do_exit(err); } - - diff --git a/arch/xtensa/kernel/vectors.S b/arch/xtensa/kernel/vectors.S index 70066e3582d..8453e6e3989 100644 --- a/arch/xtensa/kernel/vectors.S +++ b/arch/xtensa/kernel/vectors.S @@ -10,7 +10,7 @@ * Public License. See the file "COPYING" in the main directory of * this archive for more details. * - * Copyright (C) 2005 Tensilica, Inc. + * Copyright (C) 2005 - 2008 Tensilica, Inc. * * Chris Zankel <chris@zankel.net> * @@ -50,6 +50,7 @@ #include <asm/processor.h> #include <asm/page.h> #include <asm/thread_info.h> +#include <asm/vectors.h> #define WINDOW_VECTORS_SIZE 0x180 @@ -69,16 +70,19 @@ ENTRY(_UserExceptionVector) - xsr a3, EXCSAVE_1 # save a3 and get dispatch table - wsr a2, DEPC # save a2 + xsr a3, excsave1 # save a3 and get dispatch table + wsr a2, depc # save a2 l32i a2, a3, EXC_TABLE_KSTK # load kernel stack to a2 s32i a0, a2, PT_AREG0 # save a0 to ESF - rsr a0, EXCCAUSE # retrieve exception cause + rsr a0, exccause # retrieve exception cause s32i a0, a2, PT_DEPC # mark it as a regular exception addx4 a0, a0, a3 # find entry in table l32i a0, a0, EXC_TABLE_FAST_USER # load handler + xsr a3, excsave1 # restore a3 and dispatch table jx a0 +ENDPROC(_UserExceptionVector) + /* * Kernel exception vector. (Exceptions with PS.UM == 0, PS.EXCM == 0) * @@ -93,16 +97,18 @@ ENTRY(_UserExceptionVector) ENTRY(_KernelExceptionVector) - xsr a3, EXCSAVE_1 # save a3, and get dispatch table - wsr a2, DEPC # save a2 + xsr a3, excsave1 # save a3, and get dispatch table + wsr a2, depc # save a2 addi a2, a1, -16-PT_SIZE # adjust stack pointer s32i a0, a2, PT_AREG0 # save a0 to ESF - rsr a0, EXCCAUSE # retrieve exception cause + rsr a0, exccause # retrieve exception cause s32i a0, a2, PT_DEPC # mark it as a regular exception addx4 a0, a0, a3 # find entry in table l32i a0, a0, EXC_TABLE_FAST_KERNEL # load handler address + xsr a3, excsave1 # restore a3 and dispatch table jx a0 +ENDPROC(_KernelExceptionVector) /* * Double exception vector (Exceptions with PS.EXCM == 1) @@ -164,7 +170,7 @@ ENTRY(_KernelExceptionVector) * * a0: DEPC * a1: a1 - * a2: trashed, original value in EXC_TABLE_DOUBLE_A2 + * a2: trashed, original value in EXC_TABLE_DOUBLE_SAVE * a3: exctable * depc: a0 * excsave_1: a3 @@ -200,41 +206,46 @@ ENTRY(_KernelExceptionVector) .section .DoubleExceptionVector.text, "ax" .begin literal_prefix .DoubleExceptionVector + .globl _DoubleExceptionVector_WindowUnderflow + .globl _DoubleExceptionVector_WindowOverflow ENTRY(_DoubleExceptionVector) - /* Deliberately destroy excsave (don't assume it's value was valid). */ - - wsr a3, EXCSAVE_1 # save a3 + xsr a3, excsave1 + s32i a2, a3, EXC_TABLE_DOUBLE_SAVE /* Check for kernel double exception (usually fatal). */ - rsr a3, PS - _bbci.l a3, PS_UM_BIT, .Lksp + rsr a2, ps + _bbci.l a2, PS_UM_BIT, .Lksp /* Check if we are currently handling a window exception. */ /* Note: We don't need to indicate that we enter a critical section. */ - xsr a0, DEPC # get DEPC, save a0 + xsr a0, depc # get DEPC, save a0 - movi a3, XCHAL_WINDOW_VECTORS_VADDR - _bltu a0, a3, .Lfixup - addi a3, a3, WINDOW_VECTORS_SIZE - _bgeu a0, a3, .Lfixup + movi a2, WINDOW_VECTORS_VADDR + _bltu a0, a2, .Lfixup + addi a2, a2, WINDOW_VECTORS_SIZE + _bgeu a0, a2, .Lfixup /* Window overflow/underflow exception. Get stack pointer. */ - mov a3, a2 - movi a2, exc_table - l32i a2, a2, EXC_TABLE_KSTK + l32i a2, a3, EXC_TABLE_KSTK /* Check for overflow/underflow exception, jump if overflow. */ - _bbci.l a0, 6, .Lovfl - - /* a0: depc, a1: a1, a2: kstk, a3: a2, depc: a0, excsave: a3 */ + bbci.l a0, 6, _DoubleExceptionVector_WindowOverflow - /* Restart window underflow exception. + /* + * Restart window underflow exception. + * Currently: + * depc = orig a0, + * a0 = orig DEPC, + * a2 = new sp based on KSTK from exc_table + * a3 = excsave_1 + * excsave_1 = orig a3 + * * We return to the instruction in user space that caused the window * underflow exception. Therefore, we change window base to the value * before we entered the window underflow exception and prepare the @@ -242,39 +253,69 @@ ENTRY(_DoubleExceptionVector) * by changing depc (in a0). * Note: We can trash the current window frame (a0...a3) and depc! */ - - wsr a2, DEPC # save stack pointer temporarily - rsr a0, PS - extui a0, a0, PS_OWB_SHIFT, 4 - wsr a0, WINDOWBASE +_DoubleExceptionVector_WindowUnderflow: + xsr a3, excsave1 + wsr a2, depc # save stack pointer temporarily + rsr a0, ps + extui a0, a0, PS_OWB_SHIFT, PS_OWB_WIDTH + wsr a0, windowbase rsync /* We are now in the previous window frame. Save registers again. */ - xsr a2, DEPC # save a2 and get stack pointer + xsr a2, depc # save a2 and get stack pointer s32i a0, a2, PT_AREG0 - - wsr a3, EXCSAVE_1 # save a3 - movi a3, exc_table - - rsr a0, EXCCAUSE + xsr a3, excsave1 + rsr a0, exccause s32i a0, a2, PT_DEPC # mark it as a regular exception addx4 a0, a0, a3 + xsr a3, excsave1 l32i a0, a0, EXC_TABLE_FAST_USER jx a0 -.Lfixup:/* Check for a fixup handler or if we were in a critical section. */ + /* + * We only allow the ITLB miss exception if we are in kernel space. + * All other exceptions are unexpected and thus unrecoverable! + */ + +#ifdef CONFIG_MMU + .extern fast_second_level_miss_double_kernel + +.Lksp: /* a0: a0, a1: a1, a2: a2, a3: trashed, depc: depc, excsave: a3 */ + + rsr a3, exccause + beqi a3, EXCCAUSE_ITLB_MISS, 1f + addi a3, a3, -EXCCAUSE_DTLB_MISS + bnez a3, .Lunrecoverable +1: movi a3, fast_second_level_miss_double_kernel + jx a3 +#else +.equ .Lksp, .Lunrecoverable +#endif + + /* Critical! We can't handle this situation. PANIC! */ - /* a0: depc, a1: a1, a2: a2, a3: trashed, depc: a0, excsave1: a3 */ + .extern unrecoverable_exception - movi a3, exc_table - s32i a2, a3, EXC_TABLE_DOUBLE_SAVE # temporary variable +.Lunrecoverable_fixup: + l32i a2, a3, EXC_TABLE_DOUBLE_SAVE + xsr a0, depc + +.Lunrecoverable: + rsr a3, excsave1 + wsr a0, excsave1 + movi a0, unrecoverable_exception + callx0 a0 + +.Lfixup:/* Check for a fixup handler or if we were in a critical section. */ + + /* a0: depc, a1: a1, a2: trash, a3: exctable, depc: a0, excsave1: a3 */ /* Enter critical section. */ l32i a2, a3, EXC_TABLE_FIXUP s32i a3, a3, EXC_TABLE_FIXUP - beq a2, a3, .Lunrecoverable_fixup # critical! + beq a2, a3, .Lunrecoverable_fixup # critical section beqz a2, .Ldflt # no handler was registered /* a0: depc, a1: a1, a2: trash, a3: exctable, depc: a0, excsave: a3 */ @@ -283,62 +324,264 @@ ENTRY(_DoubleExceptionVector) .Ldflt: /* Get stack pointer. */ - l32i a3, a3, EXC_TABLE_DOUBLE_SAVE - addi a2, a3, -PT_USER_SIZE - -.Lovfl: /* Jump to default handlers. */ + l32i a2, a3, EXC_TABLE_DOUBLE_SAVE + addi a2, a2, -PT_USER_SIZE - /* a0: depc, a1: a1, a2: kstk, a3: a2, depc: a0, excsave: a3 */ + /* a0: depc, a1: a1, a2: kstk, a3: exctable, depc: a0, excsave: a3 */ - xsr a3, DEPC s32i a0, a2, PT_DEPC - s32i a3, a2, PT_AREG0 + l32i a0, a3, EXC_TABLE_DOUBLE_SAVE + xsr a0, depc + s32i a0, a2, PT_AREG0 - /* a0: avail, a1: a1, a2: kstk, a3: avail, depc: a2, excsave: a3 */ + /* a0: avail, a1: a1, a2: kstk, a3: exctable, depc: a2, excsave: a3 */ - movi a3, exc_table - rsr a0, EXCCAUSE + rsr a0, exccause addx4 a0, a0, a3 + xsr a3, excsave1 l32i a0, a0, EXC_TABLE_FAST_USER jx a0 /* - * We only allow the ITLB miss exception if we are in kernel space. - * All other exceptions are unexpected and thus unrecoverable! + * Restart window OVERFLOW exception. + * Currently: + * depc = orig a0, + * a0 = orig DEPC, + * a2 = new sp based on KSTK from exc_table + * a3 = EXCSAVE_1 + * excsave_1 = orig a3 + * + * We return to the instruction in user space that caused the window + * overflow exception. Therefore, we change window base to the value + * before we entered the window overflow exception and prepare the + * registers to return as if we were coming from a regular exception + * by changing DEPC (in a0). + * + * NOTE: We CANNOT trash the current window frame (a0...a3), but we + * can clobber depc. + * + * The tricky part here is that overflow8 and overflow12 handlers + * save a0, then clobber a0. To restart the handler, we have to restore + * a0 if the double exception was past the point where a0 was clobbered. + * + * To keep things simple, we take advantage of the fact all overflow + * handlers save a0 in their very first instruction. If DEPC was past + * that instruction, we can safely restore a0 from where it was saved + * on the stack. + * + * a0: depc, a1: a1, a2: kstk, a3: exc_table, depc: a0, excsave1: a3 */ +_DoubleExceptionVector_WindowOverflow: + extui a2, a0, 0, 6 # get offset into 64-byte vector handler + beqz a2, 1f # if at start of vector, don't restore -#ifdef CONFIG_MMU - .extern fast_second_level_miss_double_kernel + addi a0, a0, -128 + bbsi.l a0, 8, 1f # don't restore except for overflow 8 and 12 -.Lksp: /* a0: a0, a1: a1, a2: a2, a3: trashed, depc: depc, excsave: a3 */ + /* + * This fixup handler is for the extremely unlikely case where the + * overflow handler's reference thru a0 gets a hardware TLB refill + * that bumps out the (distinct, aliasing) TLB entry that mapped its + * prior references thru a9/a13, and where our reference now thru + * a9/a13 gets a 2nd-level miss exception (not hardware TLB refill). + */ + movi a2, window_overflow_restore_a0_fixup + s32i a2, a3, EXC_TABLE_FIXUP + l32i a2, a3, EXC_TABLE_DOUBLE_SAVE + xsr a3, excsave1 - rsr a3, EXCCAUSE - beqi a3, EXCCAUSE_ITLB_MISS, 1f - addi a3, a3, -EXCCAUSE_DTLB_MISS - bnez a3, .Lunrecoverable -1: movi a3, fast_second_level_miss_double_kernel - jx a3 -#else -.equ .Lksp, .Lunrecoverable -#endif + bbsi.l a0, 7, 2f - /* Critical! We can't handle this situation. PANIC! */ + /* + * Restore a0 as saved by _WindowOverflow8(). + */ - .extern unrecoverable_exception + l32e a0, a9, -16 + wsr a0, depc # replace the saved a0 + j 3f + +2: + /* + * Restore a0 as saved by _WindowOverflow12(). + */ + + l32e a0, a13, -16 + wsr a0, depc # replace the saved a0 +3: + xsr a3, excsave1 + movi a0, 0 + s32i a0, a3, EXC_TABLE_FIXUP + s32i a2, a3, EXC_TABLE_DOUBLE_SAVE +1: + /* + * Restore WindowBase while leaving all address registers restored. + * We have to use ROTW for this, because WSR.WINDOWBASE requires + * an address register (which would prevent restore). + * + * Window Base goes from 0 ... 7 (Module 8) + * Window Start is 8 bits; Ex: (0b1010 1010):0x55 from series of call4s + */ + + rsr a0, ps + extui a0, a0, PS_OWB_SHIFT, PS_OWB_WIDTH + rsr a2, windowbase + sub a0, a2, a0 + extui a0, a0, 0, 3 -.Lunrecoverable_fixup: l32i a2, a3, EXC_TABLE_DOUBLE_SAVE - xsr a0, DEPC + xsr a3, excsave1 + beqi a0, 1, .L1pane + beqi a0, 3, .L3pane -.Lunrecoverable: - rsr a3, EXCSAVE_1 - wsr a0, EXCSAVE_1 - movi a0, unrecoverable_exception - callx0 a0 + rsr a0, depc + rotw -2 - .end literal_prefix + /* + * We are now in the user code's original window frame. + * Process the exception as a user exception as if it was + * taken by the user code. + * + * This is similar to the user exception vector, + * except that PT_DEPC isn't set to EXCCAUSE. + */ +1: + xsr a3, excsave1 + wsr a2, depc + l32i a2, a3, EXC_TABLE_KSTK + s32i a0, a2, PT_AREG0 + rsr a0, exccause + + s32i a0, a2, PT_DEPC + +_DoubleExceptionVector_handle_exception: + addx4 a0, a0, a3 + l32i a0, a0, EXC_TABLE_FAST_USER + xsr a3, excsave1 + jx a0 + +.L1pane: + rsr a0, depc + rotw -1 + j 1b + +.L3pane: + rsr a0, depc + rotw -3 + j 1b + + +ENDPROC(_DoubleExceptionVector) + +/* + * Fixup handler for TLB miss in double exception handler for window owerflow. + * We get here with windowbase set to the window that was being spilled and + * a0 trashed. a0 bit 7 determines if this is a call8 (bit clear) or call12 + * (bit set) window. + * + * We do the following here: + * - go to the original window retaining a0 value; + * - set up exception stack to return back to appropriate a0 restore code + * (we'll need to rotate window back and there's no place to save this + * information, use different return address for that); + * - handle the exception; + * - go to the window that was being spilled; + * - set up window_overflow_restore_a0_fixup as a fixup routine; + * - reload a0; + * - restore the original window; + * - reset the default fixup routine; + * - return to user. By the time we get to this fixup handler all information + * about the conditions of the original double exception that happened in + * the window overflow handler is lost, so we just return to userspace to + * retry overflow from start. + * + * a0: value of depc, original value in depc + * a2: trashed, original value in EXC_TABLE_DOUBLE_SAVE + * a3: exctable, original value in excsave1 + */ + +ENTRY(window_overflow_restore_a0_fixup) + + rsr a0, ps + extui a0, a0, PS_OWB_SHIFT, PS_OWB_WIDTH + rsr a2, windowbase + sub a0, a2, a0 + extui a0, a0, 0, 3 + l32i a2, a3, EXC_TABLE_DOUBLE_SAVE + xsr a3, excsave1 + + _beqi a0, 1, .Lhandle_1 + _beqi a0, 3, .Lhandle_3 + + .macro overflow_fixup_handle_exception_pane n + + rsr a0, depc + rotw -\n + + xsr a3, excsave1 + wsr a2, depc + l32i a2, a3, EXC_TABLE_KSTK + s32i a0, a2, PT_AREG0 + + movi a0, .Lrestore_\n + s32i a0, a2, PT_DEPC + rsr a0, exccause + j _DoubleExceptionVector_handle_exception + + .endm + + overflow_fixup_handle_exception_pane 2 +.Lhandle_1: + overflow_fixup_handle_exception_pane 1 +.Lhandle_3: + overflow_fixup_handle_exception_pane 3 + + .macro overflow_fixup_restore_a0_pane n + + rotw \n + /* Need to preserve a0 value here to be able to handle exception + * that may occur on a0 reload from stack. It may occur because + * TLB miss handler may not be atomic and pointer to page table + * may be lost before we get here. There are no free registers, + * so we need to use EXC_TABLE_DOUBLE_SAVE area. + */ + xsr a3, excsave1 + s32i a2, a3, EXC_TABLE_DOUBLE_SAVE + movi a2, window_overflow_restore_a0_fixup + s32i a2, a3, EXC_TABLE_FIXUP + l32i a2, a3, EXC_TABLE_DOUBLE_SAVE + xsr a3, excsave1 + bbsi.l a0, 7, 1f + l32e a0, a9, -16 + j 2f +1: + l32e a0, a13, -16 +2: + rotw -\n + + .endm + +.Lrestore_2: + overflow_fixup_restore_a0_pane 2 + +.Lset_default_fixup: + xsr a3, excsave1 + s32i a2, a3, EXC_TABLE_DOUBLE_SAVE + movi a2, 0 + s32i a2, a3, EXC_TABLE_FIXUP + l32i a2, a3, EXC_TABLE_DOUBLE_SAVE + xsr a3, excsave1 + rfe + +.Lrestore_1: + overflow_fixup_restore_a0_pane 1 + j .Lset_default_fixup +.Lrestore_3: + overflow_fixup_restore_a0_pane 3 + j .Lset_default_fixup +ENDPROC(window_overflow_restore_a0_fixup) + .end literal_prefix /* * Debug interrupt vector * @@ -349,9 +592,49 @@ ENTRY(_DoubleExceptionVector) .section .DebugInterruptVector.text, "ax" ENTRY(_DebugInterruptVector) - xsr a0, EXCSAVE + XCHAL_DEBUGLEVEL + + xsr a0, SREG_EXCSAVE + XCHAL_DEBUGLEVEL jx a0 +ENDPROC(_DebugInterruptVector) + + + +/* + * Medium priority level interrupt vectors + * + * Each takes less than 16 (0x10) bytes, no literals, by placing + * the extra 8 bytes that would otherwise be required in the window + * vectors area where there is space. With relocatable vectors, + * all vectors are within ~ 4 kB range of each other, so we can + * simply jump (J) to another vector without having to use JX. + * + * common_exception code gets current IRQ level in PS.INTLEVEL + * and preserves it for the IRQ handling time. + */ + + .macro irq_entry_level level + + .if XCHAL_EXCM_LEVEL >= \level + .section .Level\level\()InterruptVector.text, "ax" +ENTRY(_Level\level\()InterruptVector) + wsr a0, excsave2 + rsr a0, epc\level + wsr a0, epc1 + movi a0, EXCCAUSE_LEVEL1_INTERRUPT + wsr a0, exccause + rsr a0, eps\level + # branch to user or kernel vector + j _SimulateUserKernelVectorException + .endif + + .endm + + irq_entry_level 2 + irq_entry_level 3 + irq_entry_level 4 + irq_entry_level 5 + irq_entry_level 6 /* Window overflow and underflow handlers. @@ -363,38 +646,61 @@ ENTRY(_DebugInterruptVector) * we try to access any page that would cause a page fault early. */ +#define ENTRY_ALIGN64(name) \ + .globl name; \ + .align 64; \ + name: + .section .WindowVectors.text, "ax" /* 4-Register Window Overflow Vector (Handler) */ - .align 64 -.global _WindowOverflow4 -_WindowOverflow4: +ENTRY_ALIGN64(_WindowOverflow4) + s32e a0, a5, -16 s32e a1, a5, -12 s32e a2, a5, -8 s32e a3, a5, -4 rfwo +ENDPROC(_WindowOverflow4) + + +#if XCHAL_EXCM_LEVEL >= 2 + /* Not a window vector - but a convenient location + * (where we know there's space) for continuation of + * medium priority interrupt dispatch code. + * On entry here, a0 contains PS, and EPC2 contains saved a0: + */ + .align 4 +_SimulateUserKernelVectorException: + addi a0, a0, (1 << PS_EXCM_BIT) + wsr a0, ps + bbsi.l a0, PS_UM_BIT, 1f # branch if user mode + rsr a0, excsave2 # restore a0 + j _KernelExceptionVector # simulate kernel vector exception +1: rsr a0, excsave2 # restore a0 + j _UserExceptionVector # simulate user vector exception +#endif + /* 4-Register Window Underflow Vector (Handler) */ - .align 64 -.global _WindowUnderflow4 -_WindowUnderflow4: +ENTRY_ALIGN64(_WindowUnderflow4) + l32e a0, a5, -16 l32e a1, a5, -12 l32e a2, a5, -8 l32e a3, a5, -4 rfwu +ENDPROC(_WindowUnderflow4) /* 8-Register Window Overflow Vector (Handler) */ - .align 64 -.global _WindowOverflow8 -_WindowOverflow8: +ENTRY_ALIGN64(_WindowOverflow8) + s32e a0, a9, -16 l32e a0, a1, -12 s32e a2, a9, -8 @@ -406,11 +712,12 @@ _WindowOverflow8: s32e a7, a0, -20 rfwo +ENDPROC(_WindowOverflow8) + /* 8-Register Window Underflow Vector (Handler) */ - .align 64 -.global _WindowUnderflow8 -_WindowUnderflow8: +ENTRY_ALIGN64(_WindowUnderflow8) + l32e a1, a9, -12 l32e a0, a9, -16 l32e a7, a1, -12 @@ -422,12 +729,12 @@ _WindowUnderflow8: l32e a7, a7, -20 rfwu +ENDPROC(_WindowUnderflow8) /* 12-Register Window Overflow Vector (Handler) */ - .align 64 -.global _WindowOverflow12 -_WindowOverflow12: +ENTRY_ALIGN64(_WindowOverflow12) + s32e a0, a13, -16 l32e a0, a1, -12 s32e a1, a13, -12 @@ -443,11 +750,12 @@ _WindowOverflow12: s32e a11, a0, -20 rfwo +ENDPROC(_WindowOverflow12) + /* 12-Register Window Underflow Vector (Handler) */ - .align 64 -.global _WindowUnderflow12 -_WindowUnderflow12: +ENTRY_ALIGN64(_WindowUnderflow12) + l32e a1, a13, -12 l32e a0, a13, -16 l32e a11, a1, -12 @@ -463,6 +771,6 @@ _WindowUnderflow12: l32e a11, a11, -20 rfwu - .text - +ENDPROC(_WindowUnderflow12) + .text diff --git a/arch/xtensa/kernel/vmlinux.lds.S b/arch/xtensa/kernel/vmlinux.lds.S index ee2e2089483..d16db6df86f 100644 --- a/arch/xtensa/kernel/vmlinux.lds.S +++ b/arch/xtensa/kernel/vmlinux.lds.S @@ -7,7 +7,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2001 - 2005 Tensilica Inc. + * Copyright (C) 2001 - 2008 Tensilica Inc. * * Chris Zankel <chris@zankel.net> * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca> @@ -18,6 +18,7 @@ #include <asm/page.h> #include <asm/thread_info.h> +#include <asm/vectors.h> #include <variant/core.h> #include <platform/hardware.h> OUTPUT_ARCH(xtensa) @@ -30,7 +31,7 @@ jiffies = jiffies_64; #endif #ifndef KERNELOFFSET -#define KERNELOFFSET 0xd0001000 +#define KERNELOFFSET 0xd0003000 #endif /* Note: In the following macros, it would be nice to specify only the @@ -134,6 +135,26 @@ SECTIONS RELOCATE_ENTRY(_WindowVectors_text, .WindowVectors.text); +#if XCHAL_EXCM_LEVEL >= 2 + RELOCATE_ENTRY(_Level2InterruptVector_text, + .Level2InterruptVector.text); +#endif +#if XCHAL_EXCM_LEVEL >= 3 + RELOCATE_ENTRY(_Level3InterruptVector_text, + .Level3InterruptVector.text); +#endif +#if XCHAL_EXCM_LEVEL >= 4 + RELOCATE_ENTRY(_Level4InterruptVector_text, + .Level4InterruptVector.text); +#endif +#if XCHAL_EXCM_LEVEL >= 5 + RELOCATE_ENTRY(_Level5InterruptVector_text, + .Level5InterruptVector.text); +#endif +#if XCHAL_EXCM_LEVEL >= 6 + RELOCATE_ENTRY(_Level6InterruptVector_text, + .Level6InterruptVector.text); +#endif RELOCATE_ENTRY(_KernelExceptionVector_text, .KernelExceptionVector.text); RELOCATE_ENTRY(_UserExceptionVector_text, @@ -144,6 +165,13 @@ SECTIONS .DoubleExceptionVector.text); RELOCATE_ENTRY(_DebugInterruptVector_text, .DebugInterruptVector.text); +#if defined(CONFIG_SMP) + RELOCATE_ENTRY(_SecondaryResetVector_literal, + .SecondaryResetVector.literal); + RELOCATE_ENTRY(_SecondaryResetVector_text, + .SecondaryResetVector.text); +#endif + __boot_reloc_table_end = ABSOLUTE(.) ; @@ -165,50 +193,111 @@ SECTIONS SECTION_VECTOR (_WindowVectors_text, .WindowVectors.text, - XCHAL_WINDOW_VECTORS_VADDR, 4, + WINDOW_VECTORS_VADDR, 4, .dummy) SECTION_VECTOR (_DebugInterruptVector_literal, .DebugInterruptVector.literal, - XCHAL_DEBUG_VECTOR_VADDR - 4, + DEBUG_VECTOR_VADDR - 4, SIZEOF(.WindowVectors.text), .WindowVectors.text) SECTION_VECTOR (_DebugInterruptVector_text, .DebugInterruptVector.text, - XCHAL_DEBUG_VECTOR_VADDR, + DEBUG_VECTOR_VADDR, 4, .DebugInterruptVector.literal) +#undef LAST +#define LAST .DebugInterruptVector.text +#if XCHAL_EXCM_LEVEL >= 2 + SECTION_VECTOR (_Level2InterruptVector_text, + .Level2InterruptVector.text, + INTLEVEL2_VECTOR_VADDR, + SIZEOF(LAST), LAST) +# undef LAST +# define LAST .Level2InterruptVector.text +#endif +#if XCHAL_EXCM_LEVEL >= 3 + SECTION_VECTOR (_Level3InterruptVector_text, + .Level3InterruptVector.text, + INTLEVEL3_VECTOR_VADDR, + SIZEOF(LAST), LAST) +# undef LAST +# define LAST .Level3InterruptVector.text +#endif +#if XCHAL_EXCM_LEVEL >= 4 + SECTION_VECTOR (_Level4InterruptVector_text, + .Level4InterruptVector.text, + INTLEVEL4_VECTOR_VADDR, + SIZEOF(LAST), LAST) +# undef LAST +# define LAST .Level4InterruptVector.text +#endif +#if XCHAL_EXCM_LEVEL >= 5 + SECTION_VECTOR (_Level5InterruptVector_text, + .Level5InterruptVector.text, + INTLEVEL5_VECTOR_VADDR, + SIZEOF(LAST), LAST) +# undef LAST +# define LAST .Level5InterruptVector.text +#endif +#if XCHAL_EXCM_LEVEL >= 6 + SECTION_VECTOR (_Level6InterruptVector_text, + .Level6InterruptVector.text, + INTLEVEL6_VECTOR_VADDR, + SIZEOF(LAST), LAST) +# undef LAST +# define LAST .Level6InterruptVector.text +#endif SECTION_VECTOR (_KernelExceptionVector_literal, .KernelExceptionVector.literal, - XCHAL_KERNEL_VECTOR_VADDR - 4, - SIZEOF(.DebugInterruptVector.text), - .DebugInterruptVector.text) + KERNEL_VECTOR_VADDR - 4, + SIZEOF(LAST), LAST) +#undef LAST SECTION_VECTOR (_KernelExceptionVector_text, .KernelExceptionVector.text, - XCHAL_KERNEL_VECTOR_VADDR, + KERNEL_VECTOR_VADDR, 4, .KernelExceptionVector.literal) SECTION_VECTOR (_UserExceptionVector_literal, .UserExceptionVector.literal, - XCHAL_USER_VECTOR_VADDR - 4, + USER_VECTOR_VADDR - 4, SIZEOF(.KernelExceptionVector.text), .KernelExceptionVector.text) SECTION_VECTOR (_UserExceptionVector_text, .UserExceptionVector.text, - XCHAL_USER_VECTOR_VADDR, + USER_VECTOR_VADDR, 4, .UserExceptionVector.literal) SECTION_VECTOR (_DoubleExceptionVector_literal, .DoubleExceptionVector.literal, - XCHAL_DOUBLEEXC_VECTOR_VADDR - 16, + DOUBLEEXC_VECTOR_VADDR - 40, SIZEOF(.UserExceptionVector.text), .UserExceptionVector.text) SECTION_VECTOR (_DoubleExceptionVector_text, .DoubleExceptionVector.text, - XCHAL_DOUBLEEXC_VECTOR_VADDR, - 32, + DOUBLEEXC_VECTOR_VADDR, + 40, .DoubleExceptionVector.literal) . = (LOADADDR( .DoubleExceptionVector.text ) + SIZEOF( .DoubleExceptionVector.text ) + 3) & ~ 3; + +#if defined(CONFIG_SMP) + + SECTION_VECTOR (_SecondaryResetVector_literal, + .SecondaryResetVector.literal, + RESET_VECTOR1_VADDR - 4, + SIZEOF(.DoubleExceptionVector.text), + .DoubleExceptionVector.text) + + SECTION_VECTOR (_SecondaryResetVector_text, + .SecondaryResetVector.text, + RESET_VECTOR1_VADDR, + 4, + .SecondaryResetVector.literal) + + . = LOADADDR(.SecondaryResetVector.text)+SIZEOF(.SecondaryResetVector.text); + +#endif + . = ALIGN(PAGE_SIZE); __init_end = .; @@ -222,16 +311,26 @@ SECTIONS . = ALIGN(0x10); .bootstrap : { *(.bootstrap.literal .bootstrap.text .bootstrap.data) } - . = ALIGN(0x1000); - __initrd_start = .; - .initrd : { *(.initrd) } - __initrd_end = .; - - .ResetVector.text XCHAL_RESET_VECTOR_VADDR : + .ResetVector.text RESET_VECTOR_VADDR : { *(.ResetVector.text) } + + /* + * This is a remapped copy of the Secondary Reset Vector Code. + * It keeps gdb in sync with the PC after switching + * to the temporary mapping used while setting up + * the V2 MMU mappings for Linux. + * + * Only debug information about this section is put in the kernel image. + */ + .SecondaryResetVector.remapped_text 0x46000000 (INFO): + { + *(.SecondaryResetVector.remapped_text) + } + + .xt.lit : { *(.xt.lit) } .xt.prop : { *(.xt.prop) } diff --git a/arch/xtensa/kernel/xtensa_ksyms.c b/arch/xtensa/kernel/xtensa_ksyms.c index c9a7c5b74a0..4d2872fd9bb 100644 --- a/arch/xtensa/kernel/xtensa_ksyms.c +++ b/arch/xtensa/kernel/xtensa_ksyms.c @@ -20,11 +20,13 @@ #include <linux/in6.h> #include <asm/uaccess.h> +#include <asm/cacheflush.h> #include <asm/checksum.h> #include <asm/dma.h> #include <asm/io.h> #include <asm/page.h> #include <asm/pgalloc.h> +#include <asm/ftrace.h> #ifdef CONFIG_BLK_DEV_FD #include <asm/floppy.h> #endif @@ -39,8 +41,11 @@ EXPORT_SYMBOL(memset); EXPORT_SYMBOL(memcpy); EXPORT_SYMBOL(memmove); +EXPORT_SYMBOL(__strncpy_user); +EXPORT_SYMBOL(clear_page); +EXPORT_SYMBOL(copy_page); -EXPORT_SYMBOL(kernel_thread); +EXPORT_SYMBOL(empty_zero_page); /* * gcc internal math functions @@ -56,6 +61,7 @@ extern unsigned int __udivsi3(unsigned int, unsigned int); extern unsigned int __umodsi3(unsigned int, unsigned int); extern unsigned long long __umoddi3(unsigned long long, unsigned long long); extern unsigned long long __udivdi3(unsigned long long, unsigned long long); +extern int __ucmpdi2(int, int); EXPORT_SYMBOL(__ashldi3); EXPORT_SYMBOL(__ashrdi3); @@ -68,11 +74,31 @@ EXPORT_SYMBOL(__udivsi3); EXPORT_SYMBOL(__umodsi3); EXPORT_SYMBOL(__udivdi3); EXPORT_SYMBOL(__umoddi3); +EXPORT_SYMBOL(__ucmpdi2); + +void __xtensa_libgcc_window_spill(void) +{ + BUG(); +} +EXPORT_SYMBOL(__xtensa_libgcc_window_spill); + +unsigned long __sync_fetch_and_and_4(unsigned long *p, unsigned long v) +{ + BUG(); +} +EXPORT_SYMBOL(__sync_fetch_and_and_4); + +unsigned long __sync_fetch_and_or_4(unsigned long *p, unsigned long v) +{ + BUG(); +} +EXPORT_SYMBOL(__sync_fetch_and_or_4); #ifdef CONFIG_NET /* * Networking support */ +EXPORT_SYMBOL(csum_partial); EXPORT_SYMBOL(csum_partial_copy_generic); #endif /* CONFIG_NET */ @@ -80,6 +106,7 @@ EXPORT_SYMBOL(csum_partial_copy_generic); * Architecture-specific symbols */ EXPORT_SYMBOL(__xtensa_copy_user); +EXPORT_SYMBOL(__invalidate_icache_range); /* * Kernel hacking ... @@ -95,3 +122,15 @@ EXPORT_SYMBOL(outsl); EXPORT_SYMBOL(insb); EXPORT_SYMBOL(insw); EXPORT_SYMBOL(insl); + +extern long common_exception_return; +EXPORT_SYMBOL(common_exception_return); + +#ifdef CONFIG_FUNCTION_TRACER +EXPORT_SYMBOL(_mcount); +#endif + +EXPORT_SYMBOL(__invalidate_dcache_range); +#if XCHAL_DCACHE_IS_WRITEBACK +EXPORT_SYMBOL(__flush_dcache_range); +#endif |