linux/arch/sparc64/kernel/trampoline.S
David S. Miller 72aff53f1f [SPARC64]: Get SUN4V SMP working.
The sibling cpu bringup is extremely fragile.  We can only
perform the most basic calls until we take over the trap
table from the firmware/hypervisor on the new cpu.

This means no accesses to %g4, %g5, %g6 since those can't be
TLB translated without our trap handlers.

In order to achieve this:

1) Change sun4v_init_mondo_queues() so that it can operate in
   several modes.

   It can allocate the queues, or install them in the current
   processor, or both.

   The boot cpu does both in it's call early on.

   Later, the boot cpu allocates the sibling cpu queue, starts
   the sibling cpu, then the sibling cpu loads them in.

2) init_cur_cpu_trap() is changed to take the current_thread_info()
   as an argument instead of reading %g6 directly on the current
   cpu.

3) Create a trampoline stack for the sibling cpus.  We do our basic
   kernel calls using this stack, which is locked into the kernel
   image, then go to our proper thread stack after taking over the
   trap table.

4) While we are in this delicate startup state, we put 0xdeadbeef
   into %g4/%g5/%g6 in order to catch accidental accesses.

5) On the final prom_set_trap_table*() call, we put &init_thread_union
   into %g6.  This is a hack to make prom_world(0) work.  All that
   wants to do is restore the %asi register using
   get_thread_current_ds().

Longer term we should just do the OBP calls to set the trap table by
hand just like we do for everything else.  This would avoid that silly
prom_world(0) issue, then we can remove the init_thread_union hack.

Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-20 01:13:22 -08:00

451 lines
10 KiB
ArmAsm

/* $Id: trampoline.S,v 1.26 2002/02/09 19:49:30 davem Exp $
* trampoline.S: Jump start slave processors on sparc64.
*
* Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
*/
#include <asm/head.h>
#include <asm/asi.h>
#include <asm/lsu.h>
#include <asm/dcr.h>
#include <asm/dcu.h>
#include <asm/pstate.h>
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/spitfire.h>
#include <asm/processor.h>
#include <asm/thread_info.h>
#include <asm/mmu.h>
#include <asm/hypervisor.h>
#include <asm/cpudata.h>
.data
.align 8
call_method:
.asciz "call-method"
.align 8
itlb_load:
.asciz "SUNW,itlb-load"
.align 8
dtlb_load:
.asciz "SUNW,dtlb-load"
/* XXX __cpuinit this thing XXX */
#define TRAMP_STACK_SIZE 1024
.align 16
tramp_stack:
.skip TRAMP_STACK_SIZE
.text
.align 8
.globl sparc64_cpu_startup, sparc64_cpu_startup_end
sparc64_cpu_startup:
BRANCH_IF_SUN4V(g1, niagara_startup)
BRANCH_IF_CHEETAH_BASE(g1, g5, cheetah_startup)
BRANCH_IF_CHEETAH_PLUS_OR_FOLLOWON(g1, g5, cheetah_plus_startup)
ba,pt %xcc, spitfire_startup
nop
cheetah_plus_startup:
/* Preserve OBP chosen DCU and DCR register settings. */
ba,pt %xcc, cheetah_generic_startup
nop
cheetah_startup:
mov DCR_BPE | DCR_RPE | DCR_SI | DCR_IFPOE | DCR_MS, %g1
wr %g1, %asr18
sethi %uhi(DCU_ME|DCU_RE|DCU_HPE|DCU_SPE|DCU_SL|DCU_WE), %g5
or %g5, %ulo(DCU_ME|DCU_RE|DCU_HPE|DCU_SPE|DCU_SL|DCU_WE), %g5
sllx %g5, 32, %g5
or %g5, DCU_DM | DCU_IM | DCU_DC | DCU_IC, %g5
stxa %g5, [%g0] ASI_DCU_CONTROL_REG
membar #Sync
/* fallthru */
cheetah_generic_startup:
mov TSB_EXTENSION_P, %g3
stxa %g0, [%g3] ASI_DMMU
stxa %g0, [%g3] ASI_IMMU
membar #Sync
mov TSB_EXTENSION_S, %g3
stxa %g0, [%g3] ASI_DMMU
membar #Sync
mov TSB_EXTENSION_N, %g3
stxa %g0, [%g3] ASI_DMMU
stxa %g0, [%g3] ASI_IMMU
membar #Sync
/* fallthru */
niagara_startup:
/* Disable STICK_INT interrupts. */
sethi %hi(0x80000000), %g5
sllx %g5, 32, %g5
wr %g5, %asr25
ba,pt %xcc, startup_continue
nop
spitfire_startup:
mov (LSU_CONTROL_IC | LSU_CONTROL_DC | LSU_CONTROL_IM | LSU_CONTROL_DM), %g1
stxa %g1, [%g0] ASI_LSU_CONTROL
membar #Sync
startup_continue:
sethi %hi(0x80000000), %g2
sllx %g2, 32, %g2
wr %g2, 0, %tick_cmpr
mov %o0, %l0
BRANCH_IF_SUN4V(g1, niagara_lock_tlb)
/* Call OBP by hand to lock KERNBASE into i/d tlbs.
* We lock 2 consequetive entries if we are 'bigkernel'.
*/
sethi %hi(prom_entry_lock), %g2
1: ldstub [%g2 + %lo(prom_entry_lock)], %g1
membar #StoreLoad | #StoreStore
brnz,pn %g1, 1b
nop
sethi %hi(p1275buf), %g2
or %g2, %lo(p1275buf), %g2
ldx [%g2 + 0x10], %l2
add %l2, -(192 + 128), %sp
flushw
sethi %hi(call_method), %g2
or %g2, %lo(call_method), %g2
stx %g2, [%sp + 2047 + 128 + 0x00]
mov 5, %g2
stx %g2, [%sp + 2047 + 128 + 0x08]
mov 1, %g2
stx %g2, [%sp + 2047 + 128 + 0x10]
sethi %hi(itlb_load), %g2
or %g2, %lo(itlb_load), %g2
stx %g2, [%sp + 2047 + 128 + 0x18]
sethi %hi(prom_mmu_ihandle_cache), %g2
lduw [%g2 + %lo(prom_mmu_ihandle_cache)], %g2
stx %g2, [%sp + 2047 + 128 + 0x20]
sethi %hi(KERNBASE), %g2
stx %g2, [%sp + 2047 + 128 + 0x28]
sethi %hi(kern_locked_tte_data), %g2
ldx [%g2 + %lo(kern_locked_tte_data)], %g2
stx %g2, [%sp + 2047 + 128 + 0x30]
mov 15, %g2
BRANCH_IF_ANY_CHEETAH(g1,g5,1f)
mov 63, %g2
1:
stx %g2, [%sp + 2047 + 128 + 0x38]
sethi %hi(p1275buf), %g2
or %g2, %lo(p1275buf), %g2
ldx [%g2 + 0x08], %o1
call %o1
add %sp, (2047 + 128), %o0
sethi %hi(bigkernel), %g2
lduw [%g2 + %lo(bigkernel)], %g2
brz,pt %g2, do_dtlb
nop
sethi %hi(call_method), %g2
or %g2, %lo(call_method), %g2
stx %g2, [%sp + 2047 + 128 + 0x00]
mov 5, %g2
stx %g2, [%sp + 2047 + 128 + 0x08]
mov 1, %g2
stx %g2, [%sp + 2047 + 128 + 0x10]
sethi %hi(itlb_load), %g2
or %g2, %lo(itlb_load), %g2
stx %g2, [%sp + 2047 + 128 + 0x18]
sethi %hi(prom_mmu_ihandle_cache), %g2
lduw [%g2 + %lo(prom_mmu_ihandle_cache)], %g2
stx %g2, [%sp + 2047 + 128 + 0x20]
sethi %hi(KERNBASE + 0x400000), %g2
stx %g2, [%sp + 2047 + 128 + 0x28]
sethi %hi(kern_locked_tte_data), %g2
ldx [%g2 + %lo(kern_locked_tte_data)], %g2
sethi %hi(0x400000), %g1
add %g2, %g1, %g2
stx %g2, [%sp + 2047 + 128 + 0x30]
mov 14, %g2
BRANCH_IF_ANY_CHEETAH(g1,g5,1f)
mov 62, %g2
1:
stx %g2, [%sp + 2047 + 128 + 0x38]
sethi %hi(p1275buf), %g2
or %g2, %lo(p1275buf), %g2
ldx [%g2 + 0x08], %o1
call %o1
add %sp, (2047 + 128), %o0
do_dtlb:
sethi %hi(call_method), %g2
or %g2, %lo(call_method), %g2
stx %g2, [%sp + 2047 + 128 + 0x00]
mov 5, %g2
stx %g2, [%sp + 2047 + 128 + 0x08]
mov 1, %g2
stx %g2, [%sp + 2047 + 128 + 0x10]
sethi %hi(dtlb_load), %g2
or %g2, %lo(dtlb_load), %g2
stx %g2, [%sp + 2047 + 128 + 0x18]
sethi %hi(prom_mmu_ihandle_cache), %g2
lduw [%g2 + %lo(prom_mmu_ihandle_cache)], %g2
stx %g2, [%sp + 2047 + 128 + 0x20]
sethi %hi(KERNBASE), %g2
stx %g2, [%sp + 2047 + 128 + 0x28]
sethi %hi(kern_locked_tte_data), %g2
ldx [%g2 + %lo(kern_locked_tte_data)], %g2
stx %g2, [%sp + 2047 + 128 + 0x30]
mov 15, %g2
BRANCH_IF_ANY_CHEETAH(g1,g5,1f)
mov 63, %g2
1:
stx %g2, [%sp + 2047 + 128 + 0x38]
sethi %hi(p1275buf), %g2
or %g2, %lo(p1275buf), %g2
ldx [%g2 + 0x08], %o1
call %o1
add %sp, (2047 + 128), %o0
sethi %hi(bigkernel), %g2
lduw [%g2 + %lo(bigkernel)], %g2
brz,pt %g2, do_unlock
nop
sethi %hi(call_method), %g2
or %g2, %lo(call_method), %g2
stx %g2, [%sp + 2047 + 128 + 0x00]
mov 5, %g2
stx %g2, [%sp + 2047 + 128 + 0x08]
mov 1, %g2
stx %g2, [%sp + 2047 + 128 + 0x10]
sethi %hi(dtlb_load), %g2
or %g2, %lo(dtlb_load), %g2
stx %g2, [%sp + 2047 + 128 + 0x18]
sethi %hi(prom_mmu_ihandle_cache), %g2
lduw [%g2 + %lo(prom_mmu_ihandle_cache)], %g2
stx %g2, [%sp + 2047 + 128 + 0x20]
sethi %hi(KERNBASE + 0x400000), %g2
stx %g2, [%sp + 2047 + 128 + 0x28]
sethi %hi(kern_locked_tte_data), %g2
ldx [%g2 + %lo(kern_locked_tte_data)], %g2
sethi %hi(0x400000), %g1
add %g2, %g1, %g2
stx %g2, [%sp + 2047 + 128 + 0x30]
mov 14, %g2
BRANCH_IF_ANY_CHEETAH(g1,g5,1f)
mov 62, %g2
1:
stx %g2, [%sp + 2047 + 128 + 0x38]
sethi %hi(p1275buf), %g2
or %g2, %lo(p1275buf), %g2
ldx [%g2 + 0x08], %o1
call %o1
add %sp, (2047 + 128), %o0
do_unlock:
sethi %hi(prom_entry_lock), %g2
stb %g0, [%g2 + %lo(prom_entry_lock)]
membar #StoreStore | #StoreLoad
ba,pt %xcc, after_lock_tlb
nop
niagara_lock_tlb:
mov HV_FAST_MMU_MAP_PERM_ADDR, %o5
sethi %hi(KERNBASE), %o0
clr %o1
sethi %hi(kern_locked_tte_data), %o2
ldx [%o2 + %lo(kern_locked_tte_data)], %o2
mov HV_MMU_IMMU, %o3
ta HV_FAST_TRAP
mov HV_FAST_MMU_MAP_PERM_ADDR, %o5
sethi %hi(KERNBASE), %o0
clr %o1
sethi %hi(kern_locked_tte_data), %o2
ldx [%o2 + %lo(kern_locked_tte_data)], %o2
mov HV_MMU_DMMU, %o3
ta HV_FAST_TRAP
sethi %hi(bigkernel), %g2
lduw [%g2 + %lo(bigkernel)], %g2
brz,pt %g2, after_lock_tlb
nop
mov HV_FAST_MMU_MAP_PERM_ADDR, %o5
sethi %hi(KERNBASE + 0x400000), %o0
clr %o1
sethi %hi(kern_locked_tte_data), %o2
ldx [%o2 + %lo(kern_locked_tte_data)], %o2
sethi %hi(0x400000), %o3
add %o2, %o3, %o2
mov HV_MMU_IMMU, %o3
ta HV_FAST_TRAP
mov HV_FAST_MMU_MAP_PERM_ADDR, %o5
sethi %hi(KERNBASE + 0x400000), %o0
clr %o1
sethi %hi(kern_locked_tte_data), %o2
ldx [%o2 + %lo(kern_locked_tte_data)], %o2
sethi %hi(0x400000), %o3
add %o2, %o3, %o2
mov HV_MMU_DMMU, %o3
ta HV_FAST_TRAP
after_lock_tlb:
wrpr %g0, (PSTATE_PRIV | PSTATE_PEF), %pstate
wr %g0, 0, %fprs
wr %g0, ASI_P, %asi
mov PRIMARY_CONTEXT, %g7
661: stxa %g0, [%g7] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g0, [%g7] ASI_MMU
.previous
membar #Sync
mov SECONDARY_CONTEXT, %g7
661: stxa %g0, [%g7] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g0, [%g7] ASI_MMU
.previous
membar #Sync
/* Everything we do here, until we properly take over the
* trap table, must be done with extreme care. We cannot
* make any references to %g6 (current thread pointer),
* %g4 (current task pointer), or %g5 (base of current cpu's
* per-cpu area) until we properly take over the trap table
* from the firmware and hypervisor.
*
* Get onto temporary stack which is in the locked kernel image.
*/
sethi %hi(tramp_stack), %g1
or %g1, %lo(tramp_stack), %g1
add %g1, TRAMP_STACK_SIZE, %g1
sub %g1, STACKFRAME_SZ + STACK_BIAS, %sp
mov 0, %fp
/* Put garbage in these registers to trap any access to them. */
set 0xdeadbeef, %g4
set 0xdeadbeef, %g5
set 0xdeadbeef, %g6
call init_irqwork_curcpu
nop
sethi %hi(tlb_type), %g3
lduw [%g3 + %lo(tlb_type)], %g2
cmp %g2, 3
bne,pt %icc, 1f
nop
call hard_smp_processor_id
nop
mov %o0, %o1
mov 0, %o0
mov 0, %o2
call sun4v_init_mondo_queues
mov 1, %o3
1: call init_cur_cpu_trap
ldx [%l0], %o0
/* Start using proper page size encodings in ctx register. */
sethi %hi(sparc64_kern_pri_context), %g3
ldx [%g3 + %lo(sparc64_kern_pri_context)], %g2
mov PRIMARY_CONTEXT, %g1
661: stxa %g2, [%g1] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g2, [%g1] ASI_MMU
.previous
membar #Sync
wrpr %g0, 0, %wstate
/* As a hack, put &init_thread_union into %g6.
* prom_world() loads from here to restore the %asi
* register.
*/
sethi %hi(init_thread_union), %g6
or %g6, %lo(init_thread_union), %g6
sethi %hi(is_sun4v), %o0
lduw [%o0 + %lo(is_sun4v)], %o0
brz,pt %o0, 1f
nop
TRAP_LOAD_TRAP_BLOCK(%g2, %g3)
add %g2, TRAP_PER_CPU_FAULT_INFO, %g2
stxa %g2, [%g0] ASI_SCRATCHPAD
/* Compute physical address:
*
* paddr = kern_base + (mmfsa_vaddr - KERNBASE)
*/
sethi %hi(KERNBASE), %g3
sub %g2, %g3, %g2
sethi %hi(kern_base), %g3
ldx [%g3 + %lo(kern_base)], %g3
add %g2, %g3, %o1
call prom_set_trap_table_sun4v
sethi %hi(sparc64_ttable_tl0), %o0
ba,pt %xcc, 2f
nop
1: call prom_set_trap_table
sethi %hi(sparc64_ttable_tl0), %o0
2: ldx [%l0], %g6
ldx [%g6 + TI_TASK], %g4
mov 1, %g5
sllx %g5, THREAD_SHIFT, %g5
sub %g5, (STACKFRAME_SZ + STACK_BIAS), %g5
add %g6, %g5, %sp
mov 0, %fp
rdpr %pstate, %o1
or %o1, PSTATE_IE, %o1
wrpr %o1, 0, %pstate
call smp_callin
nop
call cpu_idle
mov 0, %o0
call cpu_panic
nop
1: b,a,pt %xcc, 1b
.align 8
sparc64_cpu_startup_end: