asgardius/android_kernel_motorola_sm6225
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
android_kernel_motorola_sm6225/arch/sparc64/kernel/entry.S

2576 lines
57 KiB
ArmAsm
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* $Id: entry.S,v 1.144 2002/02/09 19:49:30 davem Exp $
* arch/sparc64/kernel/entry.S: Sparc64 trap low-level entry points.
*
* Copyright (C) 1995,1997 David S. Miller (davem@caip.rutgers.edu)
* Copyright (C) 1996 Eddie C. Dost (ecd@skynet.be)
* Copyright (C) 1996 Miguel de Icaza (miguel@nuclecu.unam.mx)
* Copyright (C) 1996,98,99 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
*/
#include <linux/errno.h>
#include <asm/head.h>
#include <asm/asi.h>
#include <asm/smp.h>
#include <asm/ptrace.h>
#include <asm/page.h>
#include <asm/signal.h>
#include <asm/pgtable.h>
#include <asm/processor.h>
#include <asm/visasm.h>
#include <asm/estate.h>
#include <asm/auxio.h>
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
#include <asm/sfafsr.h>
[SPARC64]: Send all device interrupts via one PIL. This is the first in a series of cleanups that will hopefully allow a seamless attempt at using the generic IRQ handling infrastructure in the Linux kernel. Define PIL_DEVICE_IRQ and vector all device interrupts through there. Get rid of the ugly pil0_dummy_{bucket,desc}, instead vector the timer interrupt directly to a specific handler since the timer interrupt is the only event that will be signaled on PIL 14. The irq_worklist is now in the per-cpu trap_block[]. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-06-20 10:20:00 +02:00
#include <asm/pil.h>
[SPARC]: Fix robust futex syscalls and wire up migrate_pages. When I added the entries for the robust futex syscall entries, I forgot to bump NR_SYSCALLS. The current situation is error-prone because NR_SYSCALLS lives in entry.S where the system call limit checks are enforced. Move the definition to asm/unistd.h in order to make this mistake much more difficult to make. And wire up sys_migrate_pages since the powerpc folks implemented the compat wrapper for us. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-06 01:51:03 +01:00
#include <asm/unistd.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#define curptr g6
.text
.align 32
/* This is trivial with the new code... */
.globl do_fpdis
do_fpdis:
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
sethi %hi(TSTATE_PEF), %g4
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %tstate, %g5
andcc %g5, %g4, %g0
be,pt %xcc, 1f
nop
rd %fprs, %g5
andcc %g5, FPRS_FEF, %g0
be,pt %xcc, 1f
nop
/* Legal state when DCR_IFPOE is set in Cheetah %dcr. */
sethi %hi(109f), %g7
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
add %g0, %g0, %g0
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
ba,a,pt %xcc, rtrap
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[SPARC64]: Add explicit register args to trap state loading macros. This, as well as making the code cleaner, allows a simplification in the TSB miss handling path. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-03 06:55:10 +01:00
1: TRAP_LOAD_THREAD_REG(%g6, %g1)
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
ldub [%g6 + TI_FPSAVED], %g5
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
wr %g0, FPRS_FEF, %fprs
andcc %g5, FPRS_FEF, %g0
be,a,pt %icc, 1f
clr %g7
ldx [%g6 + TI_GSR], %g7
1: andcc %g5, FPRS_DL, %g0
bne,pn %icc, 2f
fzero %f0
andcc %g5, FPRS_DU, %g0
bne,pn %icc, 1f
fzero %f2
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
faddd %f0, %f2, %f4
fmuld %f0, %f2, %f6
faddd %f0, %f2, %f8
fmuld %f0, %f2, %f10
faddd %f0, %f2, %f12
fmuld %f0, %f2, %f14
faddd %f0, %f2, %f16
fmuld %f0, %f2, %f18
faddd %f0, %f2, %f20
fmuld %f0, %f2, %f22
faddd %f0, %f2, %f24
fmuld %f0, %f2, %f26
faddd %f0, %f2, %f28
fmuld %f0, %f2, %f30
faddd %f0, %f2, %f32
fmuld %f0, %f2, %f34
faddd %f0, %f2, %f36
fmuld %f0, %f2, %f38
faddd %f0, %f2, %f40
fmuld %f0, %f2, %f42
faddd %f0, %f2, %f44
fmuld %f0, %f2, %f46
faddd %f0, %f2, %f48
fmuld %f0, %f2, %f50
faddd %f0, %f2, %f52
fmuld %f0, %f2, %f54
faddd %f0, %f2, %f56
fmuld %f0, %f2, %f58
b,pt %xcc, fpdis_exit2
faddd %f0, %f2, %f60
1: mov SECONDARY_CONTEXT, %g3
add %g6, TI_FPREGS + 0x80, %g1
faddd %f0, %f2, %f4
fmuld %f0, %f2, %f6
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: ldxa [%g3] ASI_DMMU, %g5
.section .sun4v_1insn_patch, "ax"
.word 661b
ldxa [%g3] ASI_MMU, %g5
.previous
[SPARC64]: Replace cheetah+ code patching with variables. Instead of code patching to handle the page size fields in the context registers, just use variables from which we get the proper values. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-05 00:23:20 +02:00
sethi %hi(sparc64_kern_sec_context), %g2
ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: stxa %g2, [%g3] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g2, [%g3] ASI_MMU
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
membar #Sync
add %g6, TI_FPREGS + 0xc0, %g2
faddd %f0, %f2, %f8
fmuld %f0, %f2, %f10
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
membar #Sync
ldda [%g1] ASI_BLK_S, %f32
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ldda [%g2] ASI_BLK_S, %f48
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
membar #Sync
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
faddd %f0, %f2, %f12
fmuld %f0, %f2, %f14
faddd %f0, %f2, %f16
fmuld %f0, %f2, %f18
faddd %f0, %f2, %f20
fmuld %f0, %f2, %f22
faddd %f0, %f2, %f24
fmuld %f0, %f2, %f26
faddd %f0, %f2, %f28
fmuld %f0, %f2, %f30
b,pt %xcc, fpdis_exit
[SPARC64]: Avoid membar instructions in delay slots. In particular, avoid membar instructions in the delay slot of a jmpl instruction. UltraSPARC-I, II, IIi, and IIe have a bug, documented in the UltraSPARC-IIi User's Manual, Appendix K, Erratum 51 The long and short of it is that if the IMU unit misses on a branch or jmpl, and there is a store buffer synchronizing membar in the delay slot, the chip can stop fetching instructions. If interrupts are enabled or some other trap is enabled, the chip will unwedge itself, but performance will suffer. We already had a workaround for this bug in a few spots, but it's better to have the entire tree sanitized for this rule. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-06-28 00:42:04 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
2: andcc %g5, FPRS_DU, %g0
bne,pt %icc, 3f
fzero %f32
mov SECONDARY_CONTEXT, %g3
fzero %f34
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: ldxa [%g3] ASI_DMMU, %g5
.section .sun4v_1insn_patch, "ax"
.word 661b
ldxa [%g3] ASI_MMU, %g5
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
add %g6, TI_FPREGS, %g1
[SPARC64]: Replace cheetah+ code patching with variables. Instead of code patching to handle the page size fields in the context registers, just use variables from which we get the proper values. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-05 00:23:20 +02:00
sethi %hi(sparc64_kern_sec_context), %g2
ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: stxa %g2, [%g3] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g2, [%g3] ASI_MMU
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
membar #Sync
add %g6, TI_FPREGS + 0x40, %g2
faddd %f32, %f34, %f36
fmuld %f32, %f34, %f38
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
membar #Sync
ldda [%g1] ASI_BLK_S, %f0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ldda [%g2] ASI_BLK_S, %f16
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
membar #Sync
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
faddd %f32, %f34, %f40
fmuld %f32, %f34, %f42
faddd %f32, %f34, %f44
fmuld %f32, %f34, %f46
faddd %f32, %f34, %f48
fmuld %f32, %f34, %f50
faddd %f32, %f34, %f52
fmuld %f32, %f34, %f54
faddd %f32, %f34, %f56
fmuld %f32, %f34, %f58
faddd %f32, %f34, %f60
fmuld %f32, %f34, %f62
ba,pt %xcc, fpdis_exit
[SPARC64]: Avoid membar instructions in delay slots. In particular, avoid membar instructions in the delay slot of a jmpl instruction. UltraSPARC-I, II, IIi, and IIe have a bug, documented in the UltraSPARC-IIi User's Manual, Appendix K, Erratum 51 The long and short of it is that if the IMU unit misses on a branch or jmpl, and there is a store buffer synchronizing membar in the delay slot, the chip can stop fetching instructions. If interrupts are enabled or some other trap is enabled, the chip will unwedge itself, but performance will suffer. We already had a workaround for this bug in a few spots, but it's better to have the entire tree sanitized for this rule. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-06-28 00:42:04 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
3: mov SECONDARY_CONTEXT, %g3
add %g6, TI_FPREGS, %g1
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: ldxa [%g3] ASI_DMMU, %g5
.section .sun4v_1insn_patch, "ax"
.word 661b
ldxa [%g3] ASI_MMU, %g5
.previous
[SPARC64]: Replace cheetah+ code patching with variables. Instead of code patching to handle the page size fields in the context registers, just use variables from which we get the proper values. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-05 00:23:20 +02:00
sethi %hi(sparc64_kern_sec_context), %g2
ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: stxa %g2, [%g3] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g2, [%g3] ASI_MMU
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
membar #Sync
mov 0x40, %g2
[SPARC64]: Fix userland FPU state corruption. We need to use stricter memory barriers around the block load and store instructions we use to save and restore the FPU register file. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-07 22:30:49 +02:00
membar #Sync
ldda [%g1] ASI_BLK_S, %f0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ldda [%g1 + %g2] ASI_BLK_S, %f16
add %g1, 0x80, %g1
ldda [%g1] ASI_BLK_S, %f32
ldda [%g1 + %g2] ASI_BLK_S, %f48
membar #Sync
fpdis_exit:
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: stxa %g5, [%g3] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g5, [%g3] ASI_MMU
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
membar #Sync
fpdis_exit2:
wr %g7, 0, %gsr
ldx [%g6 + TI_XFSR], %fsr
rdpr %tstate, %g3
or %g3, %g4, %g3 ! anal...
wrpr %g3, %tstate
wr %g0, FPRS_FEF, %fprs ! clean DU/DL bits
retry
.align 32
fp_other_bounce:
call do_fpother
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.globl do_fpother_check_fitos
.align 32
do_fpother_check_fitos:
[SPARC64]: Add explicit register args to trap state loading macros. This, as well as making the code cleaner, allows a simplification in the TSB miss handling path. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-03 06:55:10 +01:00
TRAP_LOAD_THREAD_REG(%g6, %g1)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
sethi %hi(fp_other_bounce - 4), %g7
or %g7, %lo(fp_other_bounce - 4), %g7
/* NOTE: Need to preserve %g7 until we fully commit
* to the fitos fixup.
*/
stx %fsr, [%g6 + TI_XFSR]
rdpr %tstate, %g3
andcc %g3, TSTATE_PRIV, %g0
bne,pn %xcc, do_fptrap_after_fsr
nop
ldx [%g6 + TI_XFSR], %g3
srlx %g3, 14, %g1
and %g1, 7, %g1
cmp %g1, 2 ! Unfinished FP-OP
bne,pn %xcc, do_fptrap_after_fsr
sethi %hi(1 << 23), %g1 ! Inexact
andcc %g3, %g1, %g0
bne,pn %xcc, do_fptrap_after_fsr
rdpr %tpc, %g1
lduwa [%g1] ASI_AIUP, %g3 ! This cannot ever fail
#define FITOS_MASK 0xc1f83fe0
#define FITOS_COMPARE 0x81a01880
sethi %hi(FITOS_MASK), %g1
or %g1, %lo(FITOS_MASK), %g1
and %g3, %g1, %g1
sethi %hi(FITOS_COMPARE), %g2
or %g2, %lo(FITOS_COMPARE), %g2
cmp %g1, %g2
bne,pn %xcc, do_fptrap_after_fsr
nop
std %f62, [%g6 + TI_FPREGS + (62 * 4)]
sethi %hi(fitos_table_1), %g1
and %g3, 0x1f, %g2
or %g1, %lo(fitos_table_1), %g1
sllx %g2, 2, %g2
jmpl %g1 + %g2, %g0
ba,pt %xcc, fitos_emul_continue
fitos_table_1:
fitod %f0, %f62
fitod %f1, %f62
fitod %f2, %f62
fitod %f3, %f62
fitod %f4, %f62
fitod %f5, %f62
fitod %f6, %f62
fitod %f7, %f62
fitod %f8, %f62
fitod %f9, %f62
fitod %f10, %f62
fitod %f11, %f62
fitod %f12, %f62
fitod %f13, %f62
fitod %f14, %f62
fitod %f15, %f62
fitod %f16, %f62
fitod %f17, %f62
fitod %f18, %f62
fitod %f19, %f62
fitod %f20, %f62
fitod %f21, %f62
fitod %f22, %f62
fitod %f23, %f62
fitod %f24, %f62
fitod %f25, %f62
fitod %f26, %f62
fitod %f27, %f62
fitod %f28, %f62
fitod %f29, %f62
fitod %f30, %f62
fitod %f31, %f62
fitos_emul_continue:
sethi %hi(fitos_table_2), %g1
srl %g3, 25, %g2
or %g1, %lo(fitos_table_2), %g1
and %g2, 0x1f, %g2
sllx %g2, 2, %g2
jmpl %g1 + %g2, %g0
ba,pt %xcc, fitos_emul_fini
fitos_table_2:
fdtos %f62, %f0
fdtos %f62, %f1
fdtos %f62, %f2
fdtos %f62, %f3
fdtos %f62, %f4
fdtos %f62, %f5
fdtos %f62, %f6
fdtos %f62, %f7
fdtos %f62, %f8
fdtos %f62, %f9
fdtos %f62, %f10
fdtos %f62, %f11
fdtos %f62, %f12
fdtos %f62, %f13
fdtos %f62, %f14
fdtos %f62, %f15
fdtos %f62, %f16
fdtos %f62, %f17
fdtos %f62, %f18
fdtos %f62, %f19
fdtos %f62, %f20
fdtos %f62, %f21
fdtos %f62, %f22
fdtos %f62, %f23
fdtos %f62, %f24
fdtos %f62, %f25
fdtos %f62, %f26
fdtos %f62, %f27
fdtos %f62, %f28
fdtos %f62, %f29
fdtos %f62, %f30
fdtos %f62, %f31
fitos_emul_fini:
ldd [%g6 + TI_FPREGS + (62 * 4)], %f62
done
.globl do_fptrap
.align 32
do_fptrap:
[SPARC64]: do_fptrap needs to load the thread reg into %g6. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-16 11:18:49 +01:00
TRAP_LOAD_THREAD_REG(%g6, %g1)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
stx %fsr, [%g6 + TI_XFSR]
do_fptrap_after_fsr:
ldub [%g6 + TI_FPSAVED], %g3
rd %fprs, %g1
or %g3, %g1, %g3
stb %g3, [%g6 + TI_FPSAVED]
rd %gsr, %g3
stx %g3, [%g6 + TI_GSR]
mov SECONDARY_CONTEXT, %g3
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: ldxa [%g3] ASI_DMMU, %g5
.section .sun4v_1insn_patch, "ax"
.word 661b
ldxa [%g3] ASI_MMU, %g5
.previous
[SPARC64]: Replace cheetah+ code patching with variables. Instead of code patching to handle the page size fields in the context registers, just use variables from which we get the proper values. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-05 00:23:20 +02:00
sethi %hi(sparc64_kern_sec_context), %g2
ldx [%g2 + %lo(sparc64_kern_sec_context)], %g2
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: stxa %g2, [%g3] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g2, [%g3] ASI_MMU
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
membar #Sync
add %g6, TI_FPREGS, %g2
andcc %g1, FPRS_DL, %g0
be,pn %icc, 4f
mov 0x40, %g3
stda %f0, [%g2] ASI_BLK_S
stda %f16, [%g2 + %g3] ASI_BLK_S
andcc %g1, FPRS_DU, %g0
be,pn %icc, 5f
4: add %g2, 128, %g2
stda %f32, [%g2] ASI_BLK_S
stda %f48, [%g2 + %g3] ASI_BLK_S
5: mov SECONDARY_CONTEXT, %g1
membar #Sync
[SPARC64]: Patch up mmu context register writes for sun4v. sun4v uses ASI_MMU instead of ASI_DMMU Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-08 07:13:05 +01:00
661: stxa %g5, [%g1] ASI_DMMU
.section .sun4v_1insn_patch, "ax"
.word 661b
stxa %g5, [%g1] ASI_MMU
.previous
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
membar #Sync
ba,pt %xcc, etrap
wr %g0, 0, %fprs
/* The registers for cross calls will be:
*
* DATA 0: [low 32-bits] Address of function to call, jmp to this
* [high 32-bits] MMU Context Argument 0, place in %g5
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
* DATA 1: Address Argument 1, place in %g1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
* DATA 2: Address Argument 2, place in %g7
*
* With this method we can do most of the cross-call tlb/cache
* flushing very quickly.
*/
.text
.align 32
.globl do_ivec
do_ivec:
mov 0x40, %g3
ldxa [%g3 + %g0] ASI_INTR_R, %g3
sethi %hi(KERNBASE), %g4
cmp %g3, %g4
bgeu,pn %xcc, do_ivec_xcall
srlx %g3, 32, %g5
stxa %g0, [%g0] ASI_INTR_RECEIVE
membar #Sync
[SPARC64]: Access ivector_table[] using physical addresses. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-14 06:42:46 +02:00
sethi %hi(ivector_table_pa), %g2
ldx [%g2 + %lo(ivector_table_pa)], %g2
[SPARC64]: Make IVEC pointers 64-bit. Currently we chain IVEC entries using 32-bit "pointers" because we know that the ivector_table is in the main kernel image, thus below 4GB. This uses proper 64-bit pointers instead. Whilst this bloats up the kernel image size, this sets the infrastructure necessary to significantly shrink the kernel size by using physical addresses and dynamically allocating the ivector table. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-12 11:59:40 +02:00
sllx %g3, 4, %g3
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
add %g2, %g3, %g3
[SPARC64]: Add support for IRQ pre-handlers. This allows a PCI controller to shim into IRQ delivery so that DMA queues can be drained, if necessary. If some bus specific code needs to run before an IRQ handler is invoked, the bus driver simply needs to setup the function pointer in bucket->irq_info->pre_handler and the two args bucket->irq_info->pre_handler_arg[12]. The Schizo PCI driver is converted over to use a pre-handler for the DMA write-sync processing it needs when a device is behind a PCI->PCI bus deeper than the top-level APB bridges. While we're here, clean up all of the action allocation and handling. Now, we allocate the irqaction as part of the bucket->irq_info area. There is an array of 4 irqaction (for PCI irq sharing) and a bitmask saying which entries are active. The bucket->irq_info is allocated at build_irq() time, not at request_irq() time. This simplifies request_irq() and free_irq() tremendously. The SMP dynamic IRQ retargetting code got removed in this change too. It was disabled for a few months now, and we can resurrect it in the future if we want. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-04 22:24:38 +02:00
[SPARC64]: Access ivector_table[] using physical addresses. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-14 06:42:46 +02:00
TRAP_LOAD_IRQ_WORK_PA(%g6, %g1)
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
[SPARC64]: Access ivector_table[] using physical addresses. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-14 06:42:46 +02:00
ldx [%g6], %g5
stxa %g5, [%g3] ASI_PHYS_USE_EC
stx %g3, [%g6]
[SPARC64]: Send all device interrupts via one PIL. This is the first in a series of cleanups that will hopefully allow a seamless attempt at using the generic IRQ handling infrastructure in the Linux kernel. Define PIL_DEVICE_IRQ and vector all device interrupts through there. Get rid of the ugly pil0_dummy_{bucket,desc}, instead vector the timer interrupt directly to a specific handler since the timer interrupt is the only event that will be signaled on PIL 14. The irq_worklist is now in the per-cpu trap_block[]. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-06-20 10:20:00 +02:00
wr %g0, 1 << PIL_DEVICE_IRQ, %set_softint
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
retry
do_ivec_xcall:
mov 0x50, %g1
ldxa [%g1 + %g0] ASI_INTR_R, %g1
srl %g3, 0, %g3
[SPARC64]: Add support for IRQ pre-handlers. This allows a PCI controller to shim into IRQ delivery so that DMA queues can be drained, if necessary. If some bus specific code needs to run before an IRQ handler is invoked, the bus driver simply needs to setup the function pointer in bucket->irq_info->pre_handler and the two args bucket->irq_info->pre_handler_arg[12]. The Schizo PCI driver is converted over to use a pre-handler for the DMA write-sync processing it needs when a device is behind a PCI->PCI bus deeper than the top-level APB bridges. While we're here, clean up all of the action allocation and handling. Now, we allocate the irqaction as part of the bucket->irq_info area. There is an array of 4 irqaction (for PCI irq sharing) and a bitmask saying which entries are active. The bucket->irq_info is allocated at build_irq() time, not at request_irq() time. This simplifies request_irq() and free_irq() tremendously. The SMP dynamic IRQ retargetting code got removed in this change too. It was disabled for a few months now, and we can resurrect it in the future if we want. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-04 22:24:38 +02:00
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov 0x60, %g7
ldxa [%g7 + %g0] ASI_INTR_R, %g7
stxa %g0, [%g0] ASI_INTR_RECEIVE
membar #Sync
ba,pt %xcc, 1f
nop
.align 32
1: jmpl %g3, %g0
nop
.globl getcc, setcc
getcc:
ldx [%o0 + PT_V9_TSTATE], %o1
srlx %o1, 32, %o1
and %o1, 0xf, %o1
retl
stx %o1, [%o0 + PT_V9_G1]
setcc:
ldx [%o0 + PT_V9_TSTATE], %o1
ldx [%o0 + PT_V9_G1], %o2
or %g0, %ulo(TSTATE_ICC), %o3
sllx %o3, 32, %o3
andn %o1, %o3, %o1
sllx %o2, 32, %o2
and %o2, %o3, %o2
or %o1, %o2, %o1
retl
stx %o1, [%o0 + PT_V9_TSTATE]
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
.globl utrap_trap
utrap_trap: /* %g3=handler,%g4=level */
[SPARC64]: Add explicit register args to trap state loading macros. This, as well as making the code cleaner, allows a simplification in the TSB miss handling path. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-03 06:55:10 +01:00
TRAP_LOAD_THREAD_REG(%g6, %g1)
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
ldx [%g6 + TI_UTRAPS], %g1
brnz,pt %g1, invoke_utrap
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
nop
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
ba,pt %xcc, etrap
rd %pc, %g7
mov %l4, %o1
call bad_trap
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
invoke_utrap:
sllx %g3, 3, %g3
ldx [%g1 + %g3], %g1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
save %sp, -128, %sp
rdpr %tstate, %l6
rdpr %cwp, %l7
andn %l6, TSTATE_CWP, %l6
wrpr %l6, %l7, %tstate
rdpr %tpc, %l6
rdpr %tnpc, %l7
wrpr %g1, 0, %tnpc
done
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
/* We need to carefully read the error status, ACK
* the errors, prevent recursive traps, and pass the
* information on to C code for logging.
*
* We pass the AFAR in as-is, and we encode the status
* information as described in asm-sparc64/sfafsr.h
*/
.globl __spitfire_access_error
__spitfire_access_error:
/* Disable ESTATE error reporting so that we do not
* take recursive traps and RED state the processor.
*/
stxa %g0, [%g0] ASI_ESTATE_ERROR_EN
membar #Sync
mov UDBE_UE, %g1
ldxa [%g0] ASI_AFSR, %g4 ! Get AFSR
/* __spitfire_cee_trap branches here with AFSR in %g4 and
* UDBE_CE in %g1. It only clears ESTATE_ERR_CE in the
* ESTATE Error Enable register.
*/
__spitfire_cee_trap_continue:
ldxa [%g0] ASI_AFAR, %g5 ! Get AFAR
rdpr %tt, %g3
and %g3, 0x1ff, %g3 ! Paranoia
sllx %g3, SFSTAT_TRAP_TYPE_SHIFT, %g3
or %g4, %g3, %g4
rdpr %tl, %g3
cmp %g3, 1
mov 1, %g3
bleu %xcc, 1f
sllx %g3, SFSTAT_TL_GT_ONE_SHIFT, %g3
or %g4, %g3, %g4
/* Read in the UDB error register state, clearing the
* sticky error bits as-needed. We only clear them if
* the UE bit is set. Likewise, __spitfire_cee_trap
* below will only do so if the CE bit is set.
*
* NOTE: UltraSparc-I/II have high and low UDB error
* registers, corresponding to the two UDB units
* present on those chips. UltraSparc-IIi only
* has a single UDB, called "SDB" in the manual.
* For IIi the upper UDB register always reads
* as zero so for our purposes things will just
* work with the checks below.
*/
1: ldxa [%g0] ASI_UDBH_ERROR_R, %g3
and %g3, 0x3ff, %g7 ! Paranoia
sllx %g7, SFSTAT_UDBH_SHIFT, %g7
or %g4, %g7, %g4
andcc %g3, %g1, %g3 ! UDBE_UE or UDBE_CE
be,pn %xcc, 1f
nop
stxa %g3, [%g0] ASI_UDB_ERROR_W
membar #Sync
1: mov 0x18, %g3
ldxa [%g3] ASI_UDBL_ERROR_R, %g3
and %g3, 0x3ff, %g7 ! Paranoia
sllx %g7, SFSTAT_UDBL_SHIFT, %g7
or %g4, %g7, %g4
andcc %g3, %g1, %g3 ! UDBE_UE or UDBE_CE
be,pn %xcc, 1f
nop
mov 0x18, %g7
stxa %g3, [%g7] ASI_UDB_ERROR_W
membar #Sync
1: /* Ok, now that we've latched the error state,
* clear the sticky bits in the AFSR.
*/
stxa %g4, [%g0] ASI_AFSR
membar #Sync
rdpr %tl, %g2
cmp %g2, 1
rdpr %pil, %g2
bleu,pt %xcc, 1f
wrpr %g0, 15, %pil
ba,pt %xcc, etraptl1
rd %pc, %g7
ba,pt %xcc, 2f
nop
1: ba,pt %xcc, etrap_irq
rd %pc, %g7
[SPARC64]: Add irqtrace/stacktrace/lockdep support. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-16 22:38:57 +01:00
2:
#ifdef CONFIG_TRACE_IRQFLAGS
call trace_hardirqs_off
nop
#endif
mov %l4, %o1
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
mov %l5, %o2
call spitfire_access_error
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
/* This is the trap handler entry point for ECC correctable
* errors. They are corrected, but we listen for the trap
* so that the event can be logged.
*
* Disrupting errors are either:
* 1) single-bit ECC errors during UDB reads to system
* memory
* 2) data parity errors during write-back events
*
* As far as I can make out from the manual, the CEE trap
* is only for correctable errors during memory read
* accesses by the front-end of the processor.
*
* The code below is only for trap level 1 CEE events,
* as it is the only situation where we can safely record
* and log. For trap level >1 we just clear the CE bit
* in the AFSR and return.
*
* This is just like __spiftire_access_error above, but it
* specifically handles correctable errors. If an
* uncorrectable error is indicated in the AFSR we
* will branch directly above to __spitfire_access_error
* to handle it instead. Uncorrectable therefore takes
* priority over correctable, and the error logging
* C code will notice this case by inspecting the
* trap type.
*/
.globl __spitfire_cee_trap
__spitfire_cee_trap:
ldxa [%g0] ASI_AFSR, %g4 ! Get AFSR
mov 1, %g3
sllx %g3, SFAFSR_UE_SHIFT, %g3
andcc %g4, %g3, %g0 ! Check for UE
bne,pn %xcc, __spitfire_access_error
nop
/* Ok, in this case we only have a correctable error.
* Indicate we only wish to capture that state in register
* %g1, and we only disable CE error reporting unlike UE
* handling which disables all errors.
*/
ldxa [%g0] ASI_ESTATE_ERROR_EN, %g3
andn %g3, ESTATE_ERR_CE, %g3
stxa %g3, [%g0] ASI_ESTATE_ERROR_EN
membar #Sync
/* Preserve AFSR in %g4, indicate UDB state to capture in %g1 */
ba,pt %xcc, __spitfire_cee_trap_continue
mov UDBE_CE, %g1
.globl __spitfire_data_access_exception
.globl __spitfire_data_access_exception_tl1
__spitfire_data_access_exception_tl1:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pstate, %g4
wrpr %g4, PSTATE_MG|PSTATE_AG, %pstate
mov TLB_SFSR, %g3
mov DMMU_SFAR, %g5
ldxa [%g3] ASI_DMMU, %g4 ! Get SFSR
ldxa [%g5] ASI_DMMU, %g5 ! Get SFAR
stxa %g0, [%g3] ASI_DMMU ! Clear SFSR.FaultValid bit
membar #Sync
[SPARC64]: Do not call winfix_dax blindly Verify we really are taking a data access exception trap, at TL1, from one of the window spill/fill handlers. Else call a new function, data_access_exception_tl1, to log the error. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:44:57 +02:00
rdpr %tt, %g3
cmp %g3, 0x80 ! first win spill/fill trap
blu,pn %xcc, 1f
cmp %g3, 0xff ! last win spill/fill trap
bgu,pn %xcc, 1f
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ba,pt %xcc, winfix_dax
rdpr %tpc, %g3
[SPARC64]: Do not call winfix_dax blindly Verify we really are taking a data access exception trap, at TL1, from one of the window spill/fill handlers. Else call a new function, data_access_exception_tl1, to log the error. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:44:57 +02:00
1: sethi %hi(109f), %g7
ba,pt %xcc, etraptl1
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
call spitfire_data_access_exception_tl1
[SPARC64]: Do not call winfix_dax blindly Verify we really are taking a data access exception trap, at TL1, from one of the window spill/fill handlers. Else call a new function, data_access_exception_tl1, to log the error. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:44:57 +02:00
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
[SPARC64]: Do not call winfix_dax blindly Verify we really are taking a data access exception trap, at TL1, from one of the window spill/fill handlers. Else call a new function, data_access_exception_tl1, to log the error. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:44:57 +02:00
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
__spitfire_data_access_exception:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pstate, %g4
wrpr %g4, PSTATE_MG|PSTATE_AG, %pstate
mov TLB_SFSR, %g3
mov DMMU_SFAR, %g5
ldxa [%g3] ASI_DMMU, %g4 ! Get SFSR
ldxa [%g5] ASI_DMMU, %g5 ! Get SFAR
stxa %g0, [%g3] ASI_DMMU ! Clear SFSR.FaultValid bit
membar #Sync
sethi %hi(109f), %g7
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
call spitfire_data_access_exception
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
.globl __spitfire_insn_access_exception
.globl __spitfire_insn_access_exception_tl1
__spitfire_insn_access_exception_tl1:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pstate, %g4
wrpr %g4, PSTATE_MG|PSTATE_AG, %pstate
mov TLB_SFSR, %g3
[SPARC64]: Fix trap state reading for instruction_access_exception. 1) Read ASI_IMMU SFSR not ASI_DMMU. 2) IMMU has no SFAR, read TPC instead 3) Delete old and incorrect comment about the DTLB protection trap having a dependency on the SFSR contents in order to function correctly Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:44:40 +02:00
ldxa [%g3] ASI_IMMU, %g4 ! Get SFSR
rdpr %tpc, %g5 ! IMMU has no SFAR, use TPC
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
stxa %g0, [%g3] ASI_IMMU ! Clear FaultValid bit
membar #Sync
sethi %hi(109f), %g7
ba,pt %xcc, etraptl1
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
call spitfire_insn_access_exception_tl1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
__spitfire_insn_access_exception:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pstate, %g4
wrpr %g4, PSTATE_MG|PSTATE_AG, %pstate
mov TLB_SFSR, %g3
[SPARC64]: Fix trap state reading for instruction_access_exception. 1) Read ASI_IMMU SFSR not ASI_DMMU. 2) IMMU has no SFAR, read TPC instead 3) Delete old and incorrect comment about the DTLB protection trap having a dependency on the SFSR contents in order to function correctly Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:44:40 +02:00
ldxa [%g3] ASI_IMMU, %g4 ! Get SFSR
rdpr %tpc, %g5 ! IMMU has no SFAR, use TPC
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
stxa %g0, [%g3] ASI_IMMU ! Clear FaultValid bit
membar #Sync
sethi %hi(109f), %g7
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
[SPARC64]: Revamp Spitfire error trap handling. Current uncorrectable error handling was poor enough that the processor could just loop taking the same trap over and over again. Fix things up so that we at least get a log message and perhaps even some register state. In the process, much consolidation became possible, particularly with the correctable error handler. Prefix assembler and C function names with "spitfire" to indicate that these are for Ultra-I/II/IIi/IIe only. More work is needed to make these routines robust and featureful to the level of the Ultra-III error handlers. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-29 21:45:11 +02:00
call spitfire_insn_access_exception
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* These get patched into the trap table at boot time
* once we know we have a cheetah processor.
*/
.globl cheetah_fecc_trap_vector, cheetah_fecc_trap_vector_tl1
cheetah_fecc_trap_vector:
membar #Sync
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1
andn %g1, DCU_DC | DCU_IC, %g1
stxa %g1, [%g0] ASI_DCU_CONTROL_REG
membar #Sync
sethi %hi(cheetah_fast_ecc), %g2
jmpl %g2 + %lo(cheetah_fast_ecc), %g0
mov 0, %g1
cheetah_fecc_trap_vector_tl1:
membar #Sync
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1
andn %g1, DCU_DC | DCU_IC, %g1
stxa %g1, [%g0] ASI_DCU_CONTROL_REG
membar #Sync
sethi %hi(cheetah_fast_ecc), %g2
jmpl %g2 + %lo(cheetah_fast_ecc), %g0
mov 1, %g1
.globl cheetah_cee_trap_vector, cheetah_cee_trap_vector_tl1
cheetah_cee_trap_vector:
membar #Sync
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1
andn %g1, DCU_IC, %g1
stxa %g1, [%g0] ASI_DCU_CONTROL_REG
membar #Sync
sethi %hi(cheetah_cee), %g2
jmpl %g2 + %lo(cheetah_cee), %g0
mov 0, %g1
cheetah_cee_trap_vector_tl1:
membar #Sync
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1
andn %g1, DCU_IC, %g1
stxa %g1, [%g0] ASI_DCU_CONTROL_REG
membar #Sync
sethi %hi(cheetah_cee), %g2
jmpl %g2 + %lo(cheetah_cee), %g0
mov 1, %g1
.globl cheetah_deferred_trap_vector, cheetah_deferred_trap_vector_tl1
cheetah_deferred_trap_vector:
membar #Sync
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1;
andn %g1, DCU_DC | DCU_IC, %g1;
stxa %g1, [%g0] ASI_DCU_CONTROL_REG;
membar #Sync;
sethi %hi(cheetah_deferred_trap), %g2
jmpl %g2 + %lo(cheetah_deferred_trap), %g0
mov 0, %g1
cheetah_deferred_trap_vector_tl1:
membar #Sync;
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1;
andn %g1, DCU_DC | DCU_IC, %g1;
stxa %g1, [%g0] ASI_DCU_CONTROL_REG;
membar #Sync;
sethi %hi(cheetah_deferred_trap), %g2
jmpl %g2 + %lo(cheetah_deferred_trap), %g0
mov 1, %g1
/* Cheetah+ specific traps. These are for the new I/D cache parity
* error traps. The first argument to cheetah_plus_parity_handler
* is encoded as follows:
*
* Bit0: 0=dcache,1=icache
* Bit1: 0=recoverable,1=unrecoverable
*/
.globl cheetah_plus_dcpe_trap_vector, cheetah_plus_dcpe_trap_vector_tl1
cheetah_plus_dcpe_trap_vector:
membar #Sync
sethi %hi(do_cheetah_plus_data_parity), %g7
jmpl %g7 + %lo(do_cheetah_plus_data_parity), %g0
nop
nop
nop
nop
nop
do_cheetah_plus_data_parity:
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
rdpr %pil, %g2
wrpr %g0, 15, %pil
ba,pt %xcc, etrap_irq
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rd %pc, %g7
[SPARC64]: Add irqtrace/stacktrace/lockdep support. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-16 22:38:57 +01:00
#ifdef CONFIG_TRACE_IRQFLAGS
call trace_hardirqs_off
nop
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov 0x0, %o0
call cheetah_plus_parity_error
add %sp, PTREGS_OFF, %o1
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
ba,a,pt %xcc, rtrap_irq
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
cheetah_plus_dcpe_trap_vector_tl1:
membar #Sync
wrpr PSTATE_IG | PSTATE_PEF | PSTATE_PRIV, %pstate
sethi %hi(do_dcpe_tl1), %g3
jmpl %g3 + %lo(do_dcpe_tl1), %g0
nop
nop
nop
nop
.globl cheetah_plus_icpe_trap_vector, cheetah_plus_icpe_trap_vector_tl1
cheetah_plus_icpe_trap_vector:
membar #Sync
sethi %hi(do_cheetah_plus_insn_parity), %g7
jmpl %g7 + %lo(do_cheetah_plus_insn_parity), %g0
nop
nop
nop
nop
nop
do_cheetah_plus_insn_parity:
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
rdpr %pil, %g2
wrpr %g0, 15, %pil
ba,pt %xcc, etrap_irq
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rd %pc, %g7
[SPARC64]: Add irqtrace/stacktrace/lockdep support. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-16 22:38:57 +01:00
#ifdef CONFIG_TRACE_IRQFLAGS
call trace_hardirqs_off
nop
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov 0x1, %o0
call cheetah_plus_parity_error
add %sp, PTREGS_OFF, %o1
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
ba,a,pt %xcc, rtrap_irq
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
cheetah_plus_icpe_trap_vector_tl1:
membar #Sync
wrpr PSTATE_IG | PSTATE_PEF | PSTATE_PRIV, %pstate
sethi %hi(do_icpe_tl1), %g3
jmpl %g3 + %lo(do_icpe_tl1), %g0
nop
nop
nop
nop
/* If we take one of these traps when tl >= 1, then we
* jump to interrupt globals. If some trap level above us
* was also using interrupt globals, we cannot recover.
* We may use all interrupt global registers except %g6.
*/
.globl do_dcpe_tl1, do_icpe_tl1
do_dcpe_tl1:
rdpr %tl, %g1 ! Save original trap level
mov 1, %g2 ! Setup TSTATE checking loop
sethi %hi(TSTATE_IG), %g3 ! TSTATE mask bit
1: wrpr %g2, %tl ! Set trap level to check
rdpr %tstate, %g4 ! Read TSTATE for this level
andcc %g4, %g3, %g0 ! Interrupt globals in use?
bne,a,pn %xcc, do_dcpe_tl1_fatal ! Yep, irrecoverable
wrpr %g1, %tl ! Restore original trap level
add %g2, 1, %g2 ! Next trap level
cmp %g2, %g1 ! Hit them all yet?
ble,pt %icc, 1b ! Not yet
nop
wrpr %g1, %tl ! Restore original trap level
do_dcpe_tl1_nonfatal: /* Ok we may use interrupt globals safely. */
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
sethi %hi(dcache_parity_tl1_occurred), %g2
lduw [%g2 + %lo(dcache_parity_tl1_occurred)], %g1
add %g1, 1, %g1
stw %g1, [%g2 + %lo(dcache_parity_tl1_occurred)]
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* Reset D-cache parity */
sethi %hi(1 << 16), %g1 ! D-cache size
mov (1 << 5), %g2 ! D-cache line size
sub %g1, %g2, %g1 ! Move down 1 cacheline
1: srl %g1, 14, %g3 ! Compute UTAG
membar #Sync
stxa %g3, [%g1] ASI_DCACHE_UTAG
membar #Sync
sub %g2, 8, %g3 ! 64-bit data word within line
2: membar #Sync
stxa %g0, [%g1 + %g3] ASI_DCACHE_DATA
membar #Sync
subcc %g3, 8, %g3 ! Next 64-bit data word
bge,pt %icc, 2b
nop
subcc %g1, %g2, %g1 ! Next cacheline
bge,pt %icc, 1b
nop
ba,pt %xcc, dcpe_icpe_tl1_common
nop
do_dcpe_tl1_fatal:
sethi %hi(1f), %g7
ba,pt %xcc, etraptl1
1: or %g7, %lo(1b), %g7
mov 0x2, %o0
call cheetah_plus_parity_error
add %sp, PTREGS_OFF, %o1
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
do_icpe_tl1:
rdpr %tl, %g1 ! Save original trap level
mov 1, %g2 ! Setup TSTATE checking loop
sethi %hi(TSTATE_IG), %g3 ! TSTATE mask bit
1: wrpr %g2, %tl ! Set trap level to check
rdpr %tstate, %g4 ! Read TSTATE for this level
andcc %g4, %g3, %g0 ! Interrupt globals in use?
bne,a,pn %xcc, do_icpe_tl1_fatal ! Yep, irrecoverable
wrpr %g1, %tl ! Restore original trap level
add %g2, 1, %g2 ! Next trap level
cmp %g2, %g1 ! Hit them all yet?
ble,pt %icc, 1b ! Not yet
nop
wrpr %g1, %tl ! Restore original trap level
do_icpe_tl1_nonfatal: /* Ok we may use interrupt globals safely. */
[SPARC64]: Probe D/I/E-cache config and use. At boot time, determine the D-cache, I-cache and E-cache size and line-size. Use them in cache flushes when appropriate. This change was motivated by discovering that the D-cache on UltraSparc-IIIi and later are 64K not 32K, and the flushes done by the Cheetah error handlers were assuming a 32K size. There are still some pieces of code that are hard coding things and will need to be fixed up at some point. While we're here, fix the D-cache and I-cache parity error handlers to run with interrupts disabled, and when the trap occurs at trap level > 1 log the event via a counter displayed in /proc/cpuinfo. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-09-26 09:32:17 +02:00
sethi %hi(icache_parity_tl1_occurred), %g2
lduw [%g2 + %lo(icache_parity_tl1_occurred)], %g1
add %g1, 1, %g1
stw %g1, [%g2 + %lo(icache_parity_tl1_occurred)]
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* Flush I-cache */
sethi %hi(1 << 15), %g1 ! I-cache size
mov (1 << 5), %g2 ! I-cache line size
sub %g1, %g2, %g1
1: or %g1, (2 << 3), %g3
stxa %g0, [%g3] ASI_IC_TAG
membar #Sync
subcc %g1, %g2, %g1
bge,pt %icc, 1b
nop
ba,pt %xcc, dcpe_icpe_tl1_common
nop
do_icpe_tl1_fatal:
sethi %hi(1f), %g7
ba,pt %xcc, etraptl1
1: or %g7, %lo(1b), %g7
mov 0x3, %o0
call cheetah_plus_parity_error
add %sp, PTREGS_OFF, %o1
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
dcpe_icpe_tl1_common:
/* Flush D-cache, re-enable D/I caches in DCU and finally
* retry the trapping instruction.
*/
sethi %hi(1 << 16), %g1 ! D-cache size
mov (1 << 5), %g2 ! D-cache line size
sub %g1, %g2, %g1
1: stxa %g0, [%g1] ASI_DCACHE_TAG
membar #Sync
subcc %g1, %g2, %g1
bge,pt %icc, 1b
nop
ldxa [%g0] ASI_DCU_CONTROL_REG, %g1
or %g1, (DCU_DC | DCU_IC), %g1
stxa %g1, [%g0] ASI_DCU_CONTROL_REG
membar #Sync
retry
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
/* Capture I/D/E-cache state into per-cpu error scoreboard.
*
* %g1: (TL>=0) ? 1 : 0
* %g2: scratch
* %g3: scratch
* %g4: AFSR
* %g5: AFAR
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
* %g6: unused, will have current thread ptr after etrap
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
* %g7: scratch
*/
__cheetah_log_error:
/* Put "TL1" software bit into AFSR. */
and %g1, 0x1, %g1
sllx %g1, 63, %g2
or %g4, %g2, %g4
/* Get log entry pointer for this cpu at this trap level. */
BRANCH_IF_JALAPENO(g2,g3,50f)
ldxa [%g0] ASI_SAFARI_CONFIG, %g2
srlx %g2, 17, %g2
ba,pt %xcc, 60f
and %g2, 0x3ff, %g2
50: ldxa [%g0] ASI_JBUS_CONFIG, %g2
srlx %g2, 17, %g2
and %g2, 0x1f, %g2
60: sllx %g2, 9, %g2
sethi %hi(cheetah_error_log), %g3
ldx [%g3 + %lo(cheetah_error_log)], %g3
brz,pn %g3, 80f
nop
add %g3, %g2, %g3
sllx %g1, 8, %g1
add %g3, %g1, %g1
/* %g1 holds pointer to the top of the logging scoreboard */
ldx [%g1 + 0x0], %g7
cmp %g7, -1
bne,pn %xcc, 80f
nop
stx %g4, [%g1 + 0x0]
stx %g5, [%g1 + 0x8]
add %g1, 0x10, %g1
/* %g1 now points to D-cache logging area */
set 0x3ff8, %g2 /* DC_addr mask */
and %g5, %g2, %g2 /* DC_addr bits of AFAR */
srlx %g5, 12, %g3
or %g3, 1, %g3 /* PHYS tag + valid */
10: ldxa [%g2] ASI_DCACHE_TAG, %g7
cmp %g3, %g7 /* TAG match? */
bne,pt %xcc, 13f
nop
/* Yep, what we want, capture state. */
stx %g2, [%g1 + 0x20]
stx %g7, [%g1 + 0x28]
/* A membar Sync is required before and after utag access. */
membar #Sync
ldxa [%g2] ASI_DCACHE_UTAG, %g7
membar #Sync
stx %g7, [%g1 + 0x30]
ldxa [%g2] ASI_DCACHE_SNOOP_TAG, %g7
stx %g7, [%g1 + 0x38]
clr %g3
12: ldxa [%g2 + %g3] ASI_DCACHE_DATA, %g7
stx %g7, [%g1]
add %g3, (1 << 5), %g3
cmp %g3, (4 << 5)
bl,pt %xcc, 12b
add %g1, 0x8, %g1
ba,pt %xcc, 20f
add %g1, 0x20, %g1
13: sethi %hi(1 << 14), %g7
add %g2, %g7, %g2
srlx %g2, 14, %g7
cmp %g7, 4
bl,pt %xcc, 10b
nop
add %g1, 0x40, %g1
/* %g1 now points to I-cache logging area */
20: set 0x1fe0, %g2 /* IC_addr mask */
and %g5, %g2, %g2 /* IC_addr bits of AFAR */
sllx %g2, 1, %g2 /* IC_addr[13:6]==VA[12:5] */
srlx %g5, (13 - 8), %g3 /* Make PTAG */
andn %g3, 0xff, %g3 /* Mask off undefined bits */
21: ldxa [%g2] ASI_IC_TAG, %g7
andn %g7, 0xff, %g7
cmp %g3, %g7
bne,pt %xcc, 23f
nop
/* Yep, what we want, capture state. */
stx %g2, [%g1 + 0x40]
stx %g7, [%g1 + 0x48]
add %g2, (1 << 3), %g2
ldxa [%g2] ASI_IC_TAG, %g7
add %g2, (1 << 3), %g2
stx %g7, [%g1 + 0x50]
ldxa [%g2] ASI_IC_TAG, %g7
add %g2, (1 << 3), %g2
stx %g7, [%g1 + 0x60]
ldxa [%g2] ASI_IC_TAG, %g7
stx %g7, [%g1 + 0x68]
sub %g2, (3 << 3), %g2
ldxa [%g2] ASI_IC_STAG, %g7
stx %g7, [%g1 + 0x58]
clr %g3
srlx %g2, 2, %g2
22: ldxa [%g2 + %g3] ASI_IC_INSTR, %g7
stx %g7, [%g1]
add %g3, (1 << 3), %g3
cmp %g3, (8 << 3)
bl,pt %xcc, 22b
add %g1, 0x8, %g1
ba,pt %xcc, 30f
add %g1, 0x30, %g1
23: sethi %hi(1 << 14), %g7
add %g2, %g7, %g2
srlx %g2, 14, %g7
cmp %g7, 4
bl,pt %xcc, 21b
nop
add %g1, 0x70, %g1
/* %g1 now points to E-cache logging area */
30: andn %g5, (32 - 1), %g2
stx %g2, [%g1 + 0x20]
ldxa [%g2] ASI_EC_TAG_DATA, %g7
stx %g7, [%g1 + 0x28]
ldxa [%g2] ASI_EC_R, %g0
clr %g3
31: ldxa [%g3] ASI_EC_DATA, %g7
stx %g7, [%g1 + %g3]
add %g3, 0x8, %g3
cmp %g3, 0x20
bl,pt %xcc, 31b
nop
80:
rdpr %tt, %g2
cmp %g2, 0x70
be c_fast_ecc
cmp %g2, 0x63
be c_cee
nop
ba,pt %xcc, c_deferred
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* Cheetah FECC trap handling, we get here from tl{0,1}_fecc
* in the trap table. That code has done a memory barrier
* and has disabled both the I-cache and D-cache in the DCU
* control register. The I-cache is disabled so that we may
* capture the corrupted cache line, and the D-cache is disabled
* because corrupt data may have been placed there and we don't
* want to reference it.
*
* %g1 is one if this trap occurred at %tl >= 1.
*
* Next, we turn off error reporting so that we don't recurse.
*/
.globl cheetah_fast_ecc
cheetah_fast_ecc:
ldxa [%g0] ASI_ESTATE_ERROR_EN, %g2
andn %g2, ESTATE_ERROR_NCEEN | ESTATE_ERROR_CEEN, %g2
stxa %g2, [%g0] ASI_ESTATE_ERROR_EN
membar #Sync
/* Fetch and clear AFSR/AFAR */
ldxa [%g0] ASI_AFSR, %g4
ldxa [%g0] ASI_AFAR, %g5
stxa %g4, [%g0] ASI_AFSR
membar #Sync
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
ba,pt %xcc, __cheetah_log_error
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
c_fast_ecc:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pil, %g2
wrpr %g0, 15, %pil
ba,pt %xcc, etrap_irq
rd %pc, %g7
[SPARC64]: Add irqtrace/stacktrace/lockdep support. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-16 22:38:57 +01:00
#ifdef CONFIG_TRACE_IRQFLAGS
call trace_hardirqs_off
nop
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov %l4, %o1
mov %l5, %o2
call cheetah_fecc_handler
add %sp, PTREGS_OFF, %o0
ba,a,pt %xcc, rtrap_irq
/* Our caller has disabled I-cache and performed membar Sync. */
.globl cheetah_cee
cheetah_cee:
ldxa [%g0] ASI_ESTATE_ERROR_EN, %g2
andn %g2, ESTATE_ERROR_CEEN, %g2
stxa %g2, [%g0] ASI_ESTATE_ERROR_EN
membar #Sync
/* Fetch and clear AFSR/AFAR */
ldxa [%g0] ASI_AFSR, %g4
ldxa [%g0] ASI_AFAR, %g5
stxa %g4, [%g0] ASI_AFSR
membar #Sync
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
ba,pt %xcc, __cheetah_log_error
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
c_cee:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pil, %g2
wrpr %g0, 15, %pil
ba,pt %xcc, etrap_irq
rd %pc, %g7
[SPARC64]: Add irqtrace/stacktrace/lockdep support. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-16 22:38:57 +01:00
#ifdef CONFIG_TRACE_IRQFLAGS
call trace_hardirqs_off
nop
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov %l4, %o1
mov %l5, %o2
call cheetah_cee_handler
add %sp, PTREGS_OFF, %o0
ba,a,pt %xcc, rtrap_irq
/* Our caller has disabled I-cache+D-cache and performed membar Sync. */
.globl cheetah_deferred_trap
cheetah_deferred_trap:
ldxa [%g0] ASI_ESTATE_ERROR_EN, %g2
andn %g2, ESTATE_ERROR_NCEEN | ESTATE_ERROR_CEEN, %g2
stxa %g2, [%g0] ASI_ESTATE_ERROR_EN
membar #Sync
/* Fetch and clear AFSR/AFAR */
ldxa [%g0] ASI_AFSR, %g4
ldxa [%g0] ASI_AFAR, %g5
stxa %g4, [%g0] ASI_AFSR
membar #Sync
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
ba,pt %xcc, __cheetah_log_error
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[SPARC64]: Do not expand CHEETAH_LOG_ERROR 3 times. We only need to expand this thing once, saving some text section space. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-08-31 00:11:52 +02:00
c_deferred:
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %pil, %g2
wrpr %g0, 15, %pil
ba,pt %xcc, etrap_irq
rd %pc, %g7
[SPARC64]: Add irqtrace/stacktrace/lockdep support. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-11-16 22:38:57 +01:00
#ifdef CONFIG_TRACE_IRQFLAGS
call trace_hardirqs_off
nop
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov %l4, %o1
mov %l5, %o2
call cheetah_deferred_handler
add %sp, PTREGS_OFF, %o0
ba,a,pt %xcc, rtrap_irq
.globl __do_privact
__do_privact:
mov TLB_SFSR, %g3
stxa %g0, [%g3] ASI_DMMU ! Clear FaultValid bit
membar #Sync
sethi %hi(109f), %g7
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
call do_privact
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.globl do_mna
do_mna:
rdpr %tl, %g3
cmp %g3, 1
/* Setup %g4/%g5 now as they are used in the
* winfixup code.
*/
mov TLB_SFSR, %g3
mov DMMU_SFAR, %g4
ldxa [%g4] ASI_DMMU, %g4
ldxa [%g3] ASI_DMMU, %g5
stxa %g0, [%g3] ASI_DMMU ! Clear FaultValid bit
membar #Sync
bgu,pn %icc, winfix_mna
rdpr %tpc, %g3
1: sethi %hi(109f), %g7
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
call mem_address_unaligned
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.globl do_lddfmna
do_lddfmna:
sethi %hi(109f), %g7
mov TLB_SFSR, %g4
ldxa [%g4] ASI_DMMU, %g5
stxa %g0, [%g4] ASI_DMMU ! Clear FaultValid bit
membar #Sync
mov DMMU_SFAR, %g4
ldxa [%g4] ASI_DMMU, %g4
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
call handle_lddfmna
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.globl do_stdfmna
do_stdfmna:
sethi %hi(109f), %g7
mov TLB_SFSR, %g4
ldxa [%g4] ASI_DMMU, %g5
stxa %g0, [%g4] ASI_DMMU ! Clear FaultValid bit
membar #Sync
mov DMMU_SFAR, %g4
ldxa [%g4] ASI_DMMU, %g4
ba,pt %xcc, etrap
109: or %g7, %lo(109b), %g7
mov %l4, %o1
mov %l5, %o2
call handle_stdfmna
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.globl breakpoint_trap
breakpoint_trap:
call sparc_breakpoint
add %sp, PTREGS_OFF, %o0
ba,pt %xcc, rtrap
nop
/* SunOS's execv() call only specifies the argv argument, the
* environment settings are the same as the calling processes.
*/
.globl sunos_execv
sys_execve:
sethi %hi(sparc_execve), %g1
ba,pt %xcc, execve_merge
or %g1, %lo(sparc_execve), %g1
#ifdef CONFIG_COMPAT
.globl sys_execve
sunos_execv:
stx %g0, [%sp + PTREGS_OFF + PT_V9_I2]
.globl sys32_execve
sys32_execve:
sethi %hi(sparc32_execve), %g1
or %g1, %lo(sparc32_execve), %g1
#endif
execve_merge:
flushw
jmpl %g1, %g0
add %sp, PTREGS_OFF, %o0
.globl sys_pipe, sys_sigpause, sys_nis_syscall
.globl sys_rt_sigreturn
.globl sys_ptrace
.globl sys_sigaltstack
.align 32
sys_pipe: ba,pt %xcc, sparc_pipe
add %sp, PTREGS_OFF, %o0
sys_nis_syscall:ba,pt %xcc, c_sys_nis_syscall
add %sp, PTREGS_OFF, %o0
sys_memory_ordering:
ba,pt %xcc, sparc_memory_ordering
add %sp, PTREGS_OFF, %o1
sys_sigaltstack:ba,pt %xcc, do_sigaltstack
add %i6, STACK_BIAS, %o2
#ifdef CONFIG_COMPAT
.globl sys32_sigstack
sys32_sigstack: ba,pt %xcc, do_sys32_sigstack
mov %i6, %o2
.globl sys32_sigaltstack
sys32_sigaltstack:
ba,pt %xcc, do_sys32_sigaltstack
mov %i6, %o2
#endif
.align 32
#ifdef CONFIG_COMPAT
.globl sys32_sigreturn
sys32_sigreturn:
add %sp, PTREGS_OFF, %o0
call do_sigreturn32
add %o7, 1f-.-4, %o7
nop
#endif
sys_rt_sigreturn:
add %sp, PTREGS_OFF, %o0
call do_rt_sigreturn
add %o7, 1f-.-4, %o7
nop
#ifdef CONFIG_COMPAT
.globl sys32_rt_sigreturn
sys32_rt_sigreturn:
add %sp, PTREGS_OFF, %o0
call do_rt_sigreturn32
add %o7, 1f-.-4, %o7
nop
#endif
.align 32
1: ldx [%curptr + TI_FLAGS], %l5
[SPARC64]: Add syscall auditing support. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 04:29:45 +02:00
andcc %l5, (_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT), %g0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
be,pt %icc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
add %sp, PTREGS_OFF, %o0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
call syscall_trace
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
mov 1, %o1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ba,pt %xcc, rtrap
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
nop
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* This is how fork() was meant to be done, 8 instruction entry.
*
* I questioned the following code briefly, let me clear things
* up so you must not reason on it like I did.
*
* Know the fork_kpsr etc. we use in the sparc32 port? We don't
* need it here because the only piece of window state we copy to
* the child is the CWP register. Even if the parent sleeps,
* we are safe because we stuck it into pt_regs of the parent
* so it will not change.
*
* XXX This raises the question, whether we can do the same on
* XXX sparc32 to get rid of fork_kpsr _and_ fork_kwim. The
* XXX answer is yes. We stick fork_kpsr in UREG_G0 and
* XXX fork_kwim in UREG_G1 (global registers are considered
* XXX volatile across a system call in the sparc ABI I think
* XXX if it isn't we can use regs->y instead, anyone who depends
* XXX upon the Y register being preserved across a fork deserves
* XXX to lose).
*
* In fact we should take advantage of that fact for other things
* during system calls...
*/
.globl sys_fork, sys_vfork, sys_clone, sparc_exit
.globl ret_from_syscall
.align 32
sys_vfork: /* Under Linux, vfork and fork are just special cases of clone. */
sethi %hi(0x4000 | 0x0100 | SIGCHLD), %o0
or %o0, %lo(0x4000 | 0x0100 | SIGCHLD), %o0
ba,pt %xcc, sys_clone
sys_fork: clr %o1
mov SIGCHLD, %o0
sys_clone: flushw
movrz %o1, %fp, %o1
mov 0, %o3
ba,pt %xcc, sparc_do_fork
add %sp, PTREGS_OFF, %o2
ret_from_syscall:
[SPARC64]: Move syscall success and newchild state out of thread flags. These two bits were accesses non-atomically from assembler code. So, in order to eliminate any potential races resulting from that, move these pieces of state into two bytes elsewhere in struct thread_info. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-25 04:36:26 +02:00
/* Clear current_thread_info()->new_child, and
* check performance counter stuff too.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
*/
[SPARC64]: Move syscall success and newchild state out of thread flags. These two bits were accesses non-atomically from assembler code. So, in order to eliminate any potential races resulting from that, move these pieces of state into two bytes elsewhere in struct thread_info. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-25 04:36:26 +02:00
stb %g0, [%g6 + TI_NEW_CHILD]
ldx [%g6 + TI_FLAGS], %l0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
call schedule_tail
mov %g7, %o0
andcc %l0, _TIF_PERFCTR, %g0
be,pt %icc, 1f
nop
ldx [%g6 + TI_PCR], %o7
wr %g0, %o7, %pcr
/* Blackbird errata workaround. See commentary in
* smp.c:smp_percpu_timer_interrupt() for more
* information.
*/
ba,pt %xcc, 99f
nop
.align 64
99: wr %g0, %g0, %pic
rd %pic, %g0
1: b,pt %xcc, ret_sys_call
ldx [%sp + PTREGS_OFF + PT_V9_I0], %o0
[SPARC64]: Kill hard-coded %pstate setting in sparc_exit. Just flip the bit off of whatever it's currently set to. PSTATE_IE is guarenteed to be enabled when we get here. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-01 03:34:06 +01:00
sparc_exit: rdpr %pstate, %g2
wrpr %g2, PSTATE_IE, %pstate
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
rdpr %otherwin, %g1
rdpr %cansave, %g3
add %g3, %g1, %g3
wrpr %g3, 0x0, %cansave
wrpr %g0, 0x0, %otherwin
[SPARC64]: Kill hard-coded %pstate setting in sparc_exit. Just flip the bit off of whatever it's currently set to. PSTATE_IE is guarenteed to be enabled when we get here. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-01 03:34:06 +01:00
wrpr %g2, 0x0, %pstate
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
ba,pt %xcc, sys_exit
stb %g0, [%g6 + TI_WSAVED]
linux_sparc_ni_syscall:
sethi %hi(sys_ni_syscall), %l7
b,pt %xcc, 4f
or %l7, %lo(sys_ni_syscall), %l7
linux_syscall_trace32:
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
add %sp, PTREGS_OFF, %o0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
call syscall_trace
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
clr %o1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
srl %i0, 0, %o0
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
srl %i4, 0, %o4
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
srl %i1, 0, %o1
srl %i2, 0, %o2
b,pt %xcc, 2f
srl %i3, 0, %o3
linux_syscall_trace:
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
add %sp, PTREGS_OFF, %o0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
call syscall_trace
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
clr %o1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
mov %i0, %o0
mov %i1, %o1
mov %i2, %o2
mov %i3, %o3
b,pt %xcc, 2f
mov %i4, %o4
[SPARC]: Remove SunOS and Solaris binary support. As per Documentation/feature-removal-schedule.txt Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-20 11:14:23 +02:00
/* Linux 32-bit system calls enter here... */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.align 32
.globl linux_sparc_syscall32
linux_sparc_syscall32:
/* Direct access to user regs, much faster. */
cmp %g1, NR_SYSCALLS ! IEU1 Group
bgeu,pn %xcc, linux_sparc_ni_syscall ! CTI
srl %i0, 0, %o0 ! IEU0
sll %g1, 2, %l4 ! IEU0 Group
srl %i4, 0, %o4 ! IEU1
lduw [%l7 + %l4], %l7 ! Load
srl %i1, 0, %o1 ! IEU0 Group
ldx [%curptr + TI_FLAGS], %l0 ! Load
srl %i5, 0, %o5 ! IEU1
srl %i2, 0, %o2 ! IEU0 Group
[SPARC64]: Add syscall auditing support. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 04:29:45 +02:00
andcc %l0, (_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT), %g0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
bne,pn %icc, linux_syscall_trace32 ! CTI
mov %i0, %l5 ! IEU1
call %l7 ! CTI Group brk forced
srl %i3, 0, %o3 ! IEU0
ba,a,pt %xcc, 3f
[SPARC]: Remove SunOS and Solaris binary support. As per Documentation/feature-removal-schedule.txt Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-20 11:14:23 +02:00
/* Linux native system calls enter here... */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
.align 32
sparc64: cleanup after SunOS/Solaris binary emulation removal The following cleanups are now possible: - arch/sparc64/kernel/entry.S:ret_sys_call no longer has to be global - arch/sparc64/kernel/sparc64_ksyms.c: remove no longer used prototypes Signed-off-by: Adrian Bunk <bunk@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 04:15:59 +02:00
.globl linux_sparc_syscall
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
linux_sparc_syscall:
/* Direct access to user regs, much faster. */
cmp %g1, NR_SYSCALLS ! IEU1 Group
bgeu,pn %xcc, linux_sparc_ni_syscall ! CTI
mov %i0, %o0 ! IEU0
sll %g1, 2, %l4 ! IEU0 Group
mov %i1, %o1 ! IEU1
lduw [%l7 + %l4], %l7 ! Load
4: mov %i2, %o2 ! IEU0 Group
ldx [%curptr + TI_FLAGS], %l0 ! Load
mov %i3, %o3 ! IEU1
mov %i4, %o4 ! IEU0 Group
[SPARC64]: Add syscall auditing support. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 04:29:45 +02:00
andcc %l0, (_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT), %g0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
bne,pn %icc, linux_syscall_trace ! CTI Group
mov %i0, %l5 ! IEU0
2: call %l7 ! CTI Group brk forced
mov %i5, %o5 ! IEU0
nop
3: stx %o0, [%sp + PTREGS_OFF + PT_V9_I0]
ret_sys_call:
ldx [%sp + PTREGS_OFF + PT_V9_TSTATE], %g3
ldx [%sp + PTREGS_OFF + PT_V9_TNPC], %l1 ! pc = npc
sra %o0, 0, %o0
mov %ulo(TSTATE_XCARRY | TSTATE_ICARRY), %g2
sllx %g2, 32, %g2
/* Check if force_successful_syscall_return()
* was invoked.
*/
[SPARC64]: Fix ptrace/strace Don't clobber register %l0 while checking TI_SYS_NOERROR value in syscall return path. This bug was introduced by: db7d9a4eb700be766cc9f29241483dbb1e748832 Problem narrowed down by Luis F. Ortiz and Richard Mortimer. I tried using %l2 as suggested by Luis and that works for me. Looking at the code I wonder if it makes sense to simplify the code a little bit. The following works for me but I'm not sure how to exercise the "NOERROR" codepath. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-01-09 23:35:50 +01:00
ldub [%curptr + TI_SYS_NOERROR], %l2
brnz,a,pn %l2, 80f
[SPARC64]: Move syscall success and newchild state out of thread flags. These two bits were accesses non-atomically from assembler code. So, in order to eliminate any potential races resulting from that, move these pieces of state into two bytes elsewhere in struct thread_info. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-25 04:36:26 +02:00
stb %g0, [%curptr + TI_SYS_NOERROR]
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
cmp %o0, -ERESTART_RESTARTBLOCK
bgeu,pn %xcc, 1f
[SPARC64]: Add syscall auditing support. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 04:29:45 +02:00
andcc %l0, (_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT), %l6
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
80:
/* System call success, clear Carry condition code. */
andn %g3, %g2, %g3
stx %g3, [%sp + PTREGS_OFF + PT_V9_TSTATE]
bne,pn %icc, linux_syscall_trace2
add %l1, 0x4, %l2 ! npc = npc+4
stx %l1, [%sp + PTREGS_OFF + PT_V9_TPC]
[SPARC64]: %l6 trap return handling no longer necessary. Now that we indicate the "restart system call" in the trap type field of pt_regs->magic, we don't need to set the %l6 boolean in all of the trap return paths. And we therefore don't need to pass it to do_notify_resume(). Signed-off-by: David S. Miller <davem@davemloft.net>
2008-04-24 12:15:22 +02:00
ba,pt %xcc, rtrap
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
stx %l2, [%sp + PTREGS_OFF + PT_V9_TNPC]
1:
/* System call failure, set Carry condition code.
* Also, get abs(errno) to return to the process.
*/
[SPARC64]: Add syscall auditing support. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 04:29:45 +02:00
andcc %l0, (_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT), %l6
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
sub %g0, %o0, %o0
or %g3, %g2, %g3
stx %o0, [%sp + PTREGS_OFF + PT_V9_I0]
stx %g3, [%sp + PTREGS_OFF + PT_V9_TSTATE]
bne,pn %icc, linux_syscall_trace2
add %l1, 0x4, %l2 ! npc = npc+4
stx %l1, [%sp + PTREGS_OFF + PT_V9_TPC]
b,pt %xcc, rtrap
stx %l2, [%sp + PTREGS_OFF + PT_V9_TNPC]
linux_syscall_trace2:
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
add %sp, PTREGS_OFF, %o0
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
call syscall_trace
[SPARC64]: Pass regs and entry/exit boolean to syscall_trace() Also fix a bug in 32-bit syscall tracing. We forgot to update this code when we moved over to the convention that all 32-bit syscall arguments are zero extended by default. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-07-11 01:55:48 +02:00
mov 1, %o1
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
stx %l1, [%sp + PTREGS_OFF + PT_V9_TPC]
ba,pt %xcc, rtrap
stx %l2, [%sp + PTREGS_OFF + PT_V9_TNPC]
.align 32
.globl __flushw_user
__flushw_user:
rdpr %otherwin, %g1
brz,pn %g1, 2f
clr %g2
1: save %sp, -128, %sp
rdpr %otherwin, %g1
brnz,pt %g1, 1b
add %g2, 1, %g2
1: sub %g2, 1, %g2
brnz,pt %g2, 1b
restore %g0, %g0, %g0
2: retl
nop
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
[SPARC64]: Make save_stack_trace() more efficient. Doing a 'flushw' every stack trace capture creates so much overhead that it makes lockdep next to unusable. We only care about the frame pointer chain and the function caller program counters, so flush those by hand to the stack frame. This is significantly more efficient than a 'flushw' because: 1) We only save 16 bytes per active register window to the stack. 2) This doesn't push the entire register window context of the current call chain out of the cpu, forcing register window fill traps as we return back down. Note that we can't use 'restore' and 'save' instructions to move around the register windows because that wouldn't work on Niagara processors. They optimize 'save' into a new register window by simply clearing out the registers instead of pulling them in from the on-chip register window backing store. Based upon a report by Tom Callaway. Signed-off-by: David S. Miller <davem@davemloft.net>
2008-03-25 04:06:24 +01:00
/* Flush %fp and %i7 to the stack for all register
* windows active inside of the cpu. This allows
* show_stack_trace() to avoid using an expensive
* 'flushw'.
*/
.globl stack_trace_flush
.type stack_trace_flush,#function
stack_trace_flush:
rdpr %pstate, %o0
wrpr %o0, PSTATE_IE, %pstate
rdpr %cwp, %g1
rdpr %canrestore, %g2
sub %g1, 1, %g3
1: brz,pn %g2, 2f
sub %g2, 1, %g2
wrpr %g3, %cwp
stx %fp, [%sp + STACK_BIAS + RW_V9_I6]
stx %i7, [%sp + STACK_BIAS + RW_V9_I7]
ba,pt %xcc, 1b
sub %g3, 1, %g3
2: wrpr %g1, %cwp
wrpr %o0, %pstate
retl
nop
.size stack_trace_flush,.-stack_trace_flush
[SPARC64]: Refine code sequences to get the cpu id. On uniprocessor, it's always zero for optimize that. On SMP, the jmpl to the stub kills the return address stack in the cpu branch prediction logic, so expand the code sequence inline and use a code patching section to fix things up. This also always better and explicit register selection, which will be taken advantage of in a future changeset. The hard_smp_processor_id() function is big, so do not inline it. Fix up tests for Jalapeno to also test for Serrano chips too. These tests want "jbus Ultra-IIIi" cases to match, so that is what we should test for. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:27:19 +01:00
#ifdef CONFIG_SMP
.globl hard_smp_processor_id
hard_smp_processor_id:
[SPARC64]: Fix uniprocessor IRQ targetting on SUN4V. We need to use the real hardware processor ID when targetting interrupts, not the "define to 0" thing the uniprocessor build gives us. Also, fill in the Node-ID and Agent-ID fields properly on sun4u/Safari. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-17 17:38:06 +01:00
#endif
.globl real_hard_smp_processor_id
real_hard_smp_processor_id:
[SPARC64]: Refine code sequences to get the cpu id. On uniprocessor, it's always zero for optimize that. On SMP, the jmpl to the stub kills the return address stack in the cpu branch prediction logic, so expand the code sequence inline and use a code patching section to fix things up. This also always better and explicit register selection, which will be taken advantage of in a future changeset. The hard_smp_processor_id() function is big, so do not inline it. Fix up tests for Jalapeno to also test for Serrano chips too. These tests want "jbus Ultra-IIIi" cases to match, so that is what we should test for. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:27:19 +01:00
__GET_CPUID(%o0)
[SPARC64]: Elminate all usage of hard-coded trap globals. UltraSPARC has special sets of global registers which are switched to for certain trap types. There is one set for MMU related traps, one set of Interrupt Vector processing, and another set (called the Alternate globals) for all other trap types. For what seems like forever we've hard coded the values in some of these trap registers. Some examples include: 1) Interrupt Vector global %g6 holds current processors interrupt work struct where received interrupts are managed for IRQ handler dispatch. 2) MMU global %g7 holds the base of the page tables of the currently active address space. 3) Alternate global %g6 held the current_thread_info() value. Such hardcoding has resulted in some serious issues in many areas. There are some code sequences where having another register available would help clean up the implementation. Taking traps such as cross-calls from the OBP firmware requires some trick code sequences wherein we have to save away and restore all of the special sets of global registers when we enter/exit OBP. We were also using the IMMU TSB register on SMP to hold the per-cpu area base address, which doesn't work any longer now that we actually use the TSB facility of the cpu. The implementation is pretty straight forward. One tricky bit is getting the current processor ID as that is different on different cpu variants. We use a stub with a fancy calling convention which we patch at boot time. The calling convention is that the stub is branched to and the (PC - 4) to return to is in register %g1. The cpu number is left in %g6. This stub can be invoked by using the __GET_CPUID macro. We use an array of per-cpu trap state to store the current thread and physical address of the current address space's page tables. The TRAP_LOAD_THREAD_REG loads %g6 with the current thread from this table, it uses __GET_CPUID and also clobbers %g1. TRAP_LOAD_IRQ_WORK is used by the interrupt vector processing to load the current processor's IRQ software state into %g6. It also uses __GET_CPUID and clobbers %g1. Finally, TRAP_LOAD_PGD_PHYS loads the physical address base of the current address space's page tables into %g7, it clobbers %g1 and uses __GET_CPUID. Many refinements are possible, as well as some tuning, with this stuff in place. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-27 08:24:22 +01:00
retl
nop
[SPARC64]: Move devino_to_sysino out of pci_sun4v_asm.S It is not PCI specific, it is for all system interrupts. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:02:16 +01:00
/* %o0: devhandle
* %o1: devino
*
* returns %o0: sysino
*/
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
.globl sun4v_devino_to_sysino
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_devino_to_sysino,#function
[SPARC64]: Move devino_to_sysino out of pci_sun4v_asm.S It is not PCI specific, it is for all system interrupts. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:02:16 +01:00
sun4v_devino_to_sysino:
mov HV_FAST_INTR_DEVINO2SYSINO, %o5
ta HV_FAST_TRAP
retl
mov %o1, %o0
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_devino_to_sysino, .-sun4v_devino_to_sysino
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
/* %o0: sysino
*
* returns %o0: intr_enabled (HV_INTR_{DISABLED,ENABLED})
*/
.globl sun4v_intr_getenabled
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_intr_getenabled,#function
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
sun4v_intr_getenabled:
mov HV_FAST_INTR_GETENABLED, %o5
ta HV_FAST_TRAP
retl
mov %o1, %o0
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_intr_getenabled, .-sun4v_intr_getenabled
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
/* %o0: sysino
* %o1: intr_enabled (HV_INTR_{DISABLED,ENABLED})
*/
.globl sun4v_intr_setenabled
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_intr_setenabled,#function
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
sun4v_intr_setenabled:
mov HV_FAST_INTR_SETENABLED, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_intr_setenabled, .-sun4v_intr_setenabled
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
/* %o0: sysino
*
* returns %o0: intr_state (HV_INTR_STATE_*)
*/
.globl sun4v_intr_getstate
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_intr_getstate,#function
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
sun4v_intr_getstate:
mov HV_FAST_INTR_GETSTATE, %o5
ta HV_FAST_TRAP
retl
mov %o1, %o0
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_intr_getstate, .-sun4v_intr_getstate
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
/* %o0: sysino
* %o1: intr_state (HV_INTR_STATE_*)
*/
.globl sun4v_intr_setstate
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_intr_setstate,#function
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
sun4v_intr_setstate:
mov HV_FAST_INTR_SETSTATE, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_intr_setstate, .-sun4v_intr_setstate
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
/* %o0: sysino
*
* returns %o0: cpuid
*/
.globl sun4v_intr_gettarget
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_intr_gettarget,#function
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
sun4v_intr_gettarget:
mov HV_FAST_INTR_GETTARGET, %o5
ta HV_FAST_TRAP
retl
mov %o1, %o0
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_intr_gettarget, .-sun4v_intr_gettarget
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
/* %o0: sysino
* %o1: cpuid
*/
.globl sun4v_intr_settarget
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_intr_settarget,#function
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
sun4v_intr_settarget:
mov HV_FAST_INTR_SETTARGET, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_intr_settarget, .-sun4v_intr_settarget
[SPARC64]: Implement rest of generic interrupt hypervisor calls. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-13 09:23:32 +01:00
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
/* %o0: cpuid
* %o1: pc
* %o2: rtba
* %o3: arg0
[SPARC64]: Add sun4v_cpu_qconf() hypervisor call. Call it from register_one_mondo(). Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-16 23:26:53 +01:00
*
* returns %o0: status
*/
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.globl sun4v_cpu_start
.type sun4v_cpu_start,#function
sun4v_cpu_start:
mov HV_FAST_CPU_START, %o5
[SPARC64]: Add sun4v_cpu_qconf() hypervisor call. Call it from register_one_mondo(). Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-16 23:26:53 +01:00
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_cpu_start, .-sun4v_cpu_start
[SPARC64]: Add sun4v_cpu_yield(). Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-22 00:42:09 +01:00
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
/* %o0: cpuid
*
* returns %o0: status
[SPARC64]: Add sun4v_cpu_yield(). Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-22 00:42:09 +01:00
*/
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.globl sun4v_cpu_stop
.type sun4v_cpu_stop,#function
sun4v_cpu_stop:
mov HV_FAST_CPU_STOP, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_cpu_stop, .-sun4v_cpu_stop
/* returns %o0: status */
[SPARC64]: Add sun4v_cpu_yield(). Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-22 00:42:09 +01:00
.globl sun4v_cpu_yield
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_cpu_yield, #function
[SPARC64]: Add sun4v_cpu_yield(). Signed-off-by: David S. Miller <davem@davemloft.net>
2006-02-22 00:42:09 +01:00
sun4v_cpu_yield:
mov HV_FAST_CPU_YIELD, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_cpu_yield, .-sun4v_cpu_yield
/* %o0: type
* %o1: queue paddr
* %o2: num queue entries
*
* returns %o0: status
*/
.globl sun4v_cpu_qconf
.type sun4v_cpu_qconf,#function
sun4v_cpu_qconf:
mov HV_FAST_CPU_QCONF, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_cpu_qconf, .-sun4v_cpu_qconf
[SPARC64]: Fix bugs in SUN4V cpu mondo dispatch. There were several bugs in the SUN4V cpu mondo dispatch code. In fact, if we ever got a EWOULDBLOCK or other error from the hypervisor call, we'd potentially send a cpu mondo multiple times to the same cpu and even worse we could loop until the timeout resending the same mondo over and over to such cpus. So let's bulletproof this thing as follows: 1) Implement cpu_mondo_send() and cpu_state() hypervisor calls in arch/sparc64/kernel/entry.S, add prototypes to asm/hypervisor.h 2) Don't build and update the cpulist using inline functions, this was causing the cpu mask to not get updated in the caller. 3) Disable interrupts during the entire mondo send, otherwise our cpu list and/or mondo block could get overwritten if we take an interrupt and do a cpu mondo send on the current cpu. 4) Check for all possible error return types from the cpu_mondo_send() hypervisor call. In particular: HV_EOK) Our work is done, all cpus have received the mondo. HV_CPUERROR) One or more of the cpus in the cpu list we passed to the hypervisor are in error state. Use cpu_state() calls over the entries in the cpu list to see which ones. Record them in "error_mask" and report this after we are done sending the mondo to cpus which are not in error state. HV_EWOULDBLOCK) We need to keep trying. Any other error we consider fatal, we report the event and exit immediately. 5) We only timeout if forward progress is not made. Forward progress is defined as having at least one cpu get the mondo successfully in a given cpu_mondo_send() call. Otherwise we bump a counter and delay a little. If the counter hits a limit, we signal an error and report the event. Also, smp_call_function_mask() error handling reports the number of cpus incorrectly. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-01 00:10:26 +01:00
/* %o0: num cpus in cpu list
* %o1: cpu list paddr
* %o2: mondo block paddr
*
* returns %o0: status
*/
.globl sun4v_cpu_mondo_send
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_cpu_mondo_send,#function
[SPARC64]: Fix bugs in SUN4V cpu mondo dispatch. There were several bugs in the SUN4V cpu mondo dispatch code. In fact, if we ever got a EWOULDBLOCK or other error from the hypervisor call, we'd potentially send a cpu mondo multiple times to the same cpu and even worse we could loop until the timeout resending the same mondo over and over to such cpus. So let's bulletproof this thing as follows: 1) Implement cpu_mondo_send() and cpu_state() hypervisor calls in arch/sparc64/kernel/entry.S, add prototypes to asm/hypervisor.h 2) Don't build and update the cpulist using inline functions, this was causing the cpu mask to not get updated in the caller. 3) Disable interrupts during the entire mondo send, otherwise our cpu list and/or mondo block could get overwritten if we take an interrupt and do a cpu mondo send on the current cpu. 4) Check for all possible error return types from the cpu_mondo_send() hypervisor call. In particular: HV_EOK) Our work is done, all cpus have received the mondo. HV_CPUERROR) One or more of the cpus in the cpu list we passed to the hypervisor are in error state. Use cpu_state() calls over the entries in the cpu list to see which ones. Record them in "error_mask" and report this after we are done sending the mondo to cpus which are not in error state. HV_EWOULDBLOCK) We need to keep trying. Any other error we consider fatal, we report the event and exit immediately. 5) We only timeout if forward progress is not made. Forward progress is defined as having at least one cpu get the mondo successfully in a given cpu_mondo_send() call. Otherwise we bump a counter and delay a little. If the counter hits a limit, we signal an error and report the event. Also, smp_call_function_mask() error handling reports the number of cpus incorrectly. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-01 00:10:26 +01:00
sun4v_cpu_mondo_send:
mov HV_FAST_CPU_MONDO_SEND, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_cpu_mondo_send, .-sun4v_cpu_mondo_send
[SPARC64]: Fix bugs in SUN4V cpu mondo dispatch. There were several bugs in the SUN4V cpu mondo dispatch code. In fact, if we ever got a EWOULDBLOCK or other error from the hypervisor call, we'd potentially send a cpu mondo multiple times to the same cpu and even worse we could loop until the timeout resending the same mondo over and over to such cpus. So let's bulletproof this thing as follows: 1) Implement cpu_mondo_send() and cpu_state() hypervisor calls in arch/sparc64/kernel/entry.S, add prototypes to asm/hypervisor.h 2) Don't build and update the cpulist using inline functions, this was causing the cpu mask to not get updated in the caller. 3) Disable interrupts during the entire mondo send, otherwise our cpu list and/or mondo block could get overwritten if we take an interrupt and do a cpu mondo send on the current cpu. 4) Check for all possible error return types from the cpu_mondo_send() hypervisor call. In particular: HV_EOK) Our work is done, all cpus have received the mondo. HV_CPUERROR) One or more of the cpus in the cpu list we passed to the hypervisor are in error state. Use cpu_state() calls over the entries in the cpu list to see which ones. Record them in "error_mask" and report this after we are done sending the mondo to cpus which are not in error state. HV_EWOULDBLOCK) We need to keep trying. Any other error we consider fatal, we report the event and exit immediately. 5) We only timeout if forward progress is not made. Forward progress is defined as having at least one cpu get the mondo successfully in a given cpu_mondo_send() call. Otherwise we bump a counter and delay a little. If the counter hits a limit, we signal an error and report the event. Also, smp_call_function_mask() error handling reports the number of cpus incorrectly. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-01 00:10:26 +01:00
/* %o0: CPU ID
*
* returns %o0: -status if status non-zero, else
* %o0: cpu state as HV_CPU_STATE_*
*/
.globl sun4v_cpu_state
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_cpu_state,#function
[SPARC64]: Fix bugs in SUN4V cpu mondo dispatch. There were several bugs in the SUN4V cpu mondo dispatch code. In fact, if we ever got a EWOULDBLOCK or other error from the hypervisor call, we'd potentially send a cpu mondo multiple times to the same cpu and even worse we could loop until the timeout resending the same mondo over and over to such cpus. So let's bulletproof this thing as follows: 1) Implement cpu_mondo_send() and cpu_state() hypervisor calls in arch/sparc64/kernel/entry.S, add prototypes to asm/hypervisor.h 2) Don't build and update the cpulist using inline functions, this was causing the cpu mask to not get updated in the caller. 3) Disable interrupts during the entire mondo send, otherwise our cpu list and/or mondo block could get overwritten if we take an interrupt and do a cpu mondo send on the current cpu. 4) Check for all possible error return types from the cpu_mondo_send() hypervisor call. In particular: HV_EOK) Our work is done, all cpus have received the mondo. HV_CPUERROR) One or more of the cpus in the cpu list we passed to the hypervisor are in error state. Use cpu_state() calls over the entries in the cpu list to see which ones. Record them in "error_mask" and report this after we are done sending the mondo to cpus which are not in error state. HV_EWOULDBLOCK) We need to keep trying. Any other error we consider fatal, we report the event and exit immediately. 5) We only timeout if forward progress is not made. Forward progress is defined as having at least one cpu get the mondo successfully in a given cpu_mondo_send() call. Otherwise we bump a counter and delay a little. If the counter hits a limit, we signal an error and report the event. Also, smp_call_function_mask() error handling reports the number of cpus incorrectly. Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-01 00:10:26 +01:00
sun4v_cpu_state:
mov HV_FAST_CPU_STATE, %o5
ta HV_FAST_TRAP
brnz,pn %o0, 1f
sub %g0, %o0, %o0
mov %o1, %o0
1: retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_cpu_state, .-sun4v_cpu_state
/* %o0: virtual address
* %o1: must be zero
* %o2: TTE
* %o3: HV_MMU_* flags
*
* returns %o0: status
*/
.globl sun4v_mmu_map_perm_addr
.type sun4v_mmu_map_perm_addr,#function
sun4v_mmu_map_perm_addr:
mov HV_FAST_MMU_MAP_PERM_ADDR, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_mmu_map_perm_addr, .-sun4v_mmu_map_perm_addr
/* %o0: number of TSB descriptions
* %o1: TSB descriptions real address
*
* returns %o0: status
*/
.globl sun4v_mmu_tsb_ctx0
.type sun4v_mmu_tsb_ctx0,#function
sun4v_mmu_tsb_ctx0:
mov HV_FAST_MMU_TSB_CTX0, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_mmu_tsb_ctx0, .-sun4v_mmu_tsb_ctx0
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
/* %o0: API group number
* %o1: pointer to unsigned long major number storage
* %o2: pointer to unsigned long minor number storage
*
* returns %o0: status
*/
.globl sun4v_get_version
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_get_version,#function
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
sun4v_get_version:
mov HV_CORE_GET_VER, %o5
mov %o1, %o3
mov %o2, %o4
ta HV_CORE_TRAP
stx %o1, [%o3]
retl
stx %o2, [%o4]
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_get_version, .-sun4v_get_version
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
/* %o0: API group number
* %o1: desired major number
* %o2: desired minor number
* %o3: pointer to unsigned long actual minor number storage
*
* returns %o0: status
*/
.globl sun4v_set_version
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_set_version,#function
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
sun4v_set_version:
mov HV_CORE_SET_VER, %o5
mov %o3, %o4
ta HV_CORE_TRAP
retl
stx %o1, [%o4]
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_set_version, .-sun4v_set_version
/* %o0: pointer to unsigned long time
*
* returns %o0: status
*/
.globl sun4v_tod_get
.type sun4v_tod_get,#function
sun4v_tod_get:
mov %o0, %o4
mov HV_FAST_TOD_GET, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
.size sun4v_tod_get, .-sun4v_tod_get
/* %o0: time
*
* returns %o0: status
*/
.globl sun4v_tod_set
.type sun4v_tod_set,#function
sun4v_tod_set:
mov HV_FAST_TOD_SET, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_tod_set, .-sun4v_tod_set
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
/* %o0: pointer to unsigned long status
*
* returns %o0: signed character
*/
.globl sun4v_con_getchar
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_con_getchar,#function
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
sun4v_con_getchar:
mov %o0, %o4
mov HV_FAST_CONS_GETCHAR, %o5
clr %o0
clr %o1
ta HV_FAST_TRAP
stx %o0, [%o4]
retl
sra %o1, 0, %o0
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_con_getchar, .-sun4v_con_getchar
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
/* %o0: signed long character
*
* returns %o0: status
*/
.globl sun4v_con_putchar
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_con_putchar,#function
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
sun4v_con_putchar:
mov HV_FAST_CONS_PUTCHAR, %o5
ta HV_FAST_TRAP
retl
sra %o0, 0, %o0
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_con_putchar, .-sun4v_con_putchar
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
/* %o0: buffer real address
* %o1: buffer size
* %o2: pointer to unsigned long bytes_read
*
* returns %o0: status
*/
.globl sun4v_con_read
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_con_read,#function
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
sun4v_con_read:
mov %o2, %o4
mov HV_FAST_CONS_READ, %o5
ta HV_FAST_TRAP
brnz %o0, 1f
cmp %o1, -1 /* break */
be,a,pn %icc, 1f
mov %o1, %o0
cmp %o1, -2 /* hup */
be,a,pn %icc, 1f
mov %o1, %o0
stx %o1, [%o4]
1: retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_con_read, .-sun4v_con_read
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
/* %o0: buffer real address
* %o1: buffer size
* %o2: pointer to unsigned long bytes_written
*
* returns %o0: status
*/
.globl sun4v_con_write
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_con_write,#function
[SPARC64]: Add hypervisor API negotiation and fix console bugs. Hypervisor interfaces need to be negotiated in order to use some API calls reliably. So add a small set of interfaces to request API versions and query current settings. This allows us to fix some bugs in the hypervisor console: 1) If we can negotiate API group CORE of at least major 1 minor 1 we can use con_read and con_write which can improve console performance quite a bit. 2) When we do a console write request, we should hold the spinlock around the whole request, not a byte at a time. What would happen is that it's easy for output from different cpus to get mixed with each other. 3) Use consistent udelay() based polling, udelay(1) each loop with a limit of 1000 polls to handle stuck hypervisor console. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-16 02:03:54 +02:00
sun4v_con_write:
mov %o2, %o4
mov HV_FAST_CONS_WRITE, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_con_write, .-sun4v_con_write
[SPARC64]: Report proper system soft state to the hypervisor. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-25 09:37:12 +02:00
/* %o0: soft state
* %o1: address of description string
*
* returns %o0: status
*/
.globl sun4v_mach_set_soft_state
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_mach_set_soft_state,#function
[SPARC64]: Report proper system soft state to the hypervisor. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-25 09:37:12 +02:00
sun4v_mach_set_soft_state:
mov HV_FAST_MACH_SET_SOFT_STATE, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_mach_set_soft_state, .-sun4v_mach_set_soft_state
[SPARC64]: Use machine description and OBP properly for cpu probing. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-26 00:49:59 +02:00
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
/* %o0: exit code
*
* Does not return.
*/
.globl sun4v_mach_exit
.type sun4v_mach_exit,#function
sun4v_mach_exit:
mov HV_FAST_MACH_EXIT, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_mach_exit, .-sun4v_mach_exit
/* %o0: buffer real address
* %o1: buffer length
* %o2: pointer to unsigned long real_buf_len
*
* returns %o0: status
*/
[SPARC64]: Use machine description and OBP properly for cpu probing. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-26 00:49:59 +02:00
.globl sun4v_mach_desc
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.type sun4v_mach_desc,#function
[SPARC64]: Use machine description and OBP properly for cpu probing. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-26 00:49:59 +02:00
sun4v_mach_desc:
mov %o2, %o4
mov HV_FAST_MACH_DESC, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
.size sun4v_mach_desc, .-sun4v_mach_desc
/* %o0: new timeout in milliseconds
* %o1: pointer to unsigned long orig_timeout
*
* returns %o0: status
*/
.globl sun4v_mach_set_watchdog
.type sun4v_mach_set_watchdog,#function
sun4v_mach_set_watchdog:
mov %o1, %o4
mov HV_FAST_MACH_SET_WATCHDOG, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
.size sun4v_mach_set_watchdog, .-sun4v_mach_set_watchdog
/* No inputs and does not return. */
.globl sun4v_mach_sir
.type sun4v_mach_sir,#function
sun4v_mach_sir:
mov %o1, %o4
mov HV_FAST_MACH_SIR, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
.size sun4v_mach_sir, .-sun4v_mach_sir
/* %o0: channel
* %o1: ra
* %o2: num_entries
*
* returns %o0: status
*/
.globl sun4v_ldc_tx_qconf
.type sun4v_ldc_tx_qconf,#function
sun4v_ldc_tx_qconf:
mov HV_FAST_LDC_TX_QCONF, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_tx_qconf, .-sun4v_ldc_tx_qconf
/* %o0: channel
* %o1: pointer to unsigned long ra
* %o2: pointer to unsigned long num_entries
*
* returns %o0: status
*/
.globl sun4v_ldc_tx_qinfo
.type sun4v_ldc_tx_qinfo,#function
sun4v_ldc_tx_qinfo:
mov %o1, %g1
mov %o2, %g2
mov HV_FAST_LDC_TX_QINFO, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
stx %o2, [%g2]
retl
nop
.size sun4v_ldc_tx_qinfo, .-sun4v_ldc_tx_qinfo
/* %o0: channel
* %o1: pointer to unsigned long head_off
* %o2: pointer to unsigned long tail_off
* %o2: pointer to unsigned long chan_state
*
* returns %o0: status
*/
.globl sun4v_ldc_tx_get_state
.type sun4v_ldc_tx_get_state,#function
sun4v_ldc_tx_get_state:
mov %o1, %g1
mov %o2, %g2
mov %o3, %g3
mov HV_FAST_LDC_TX_GET_STATE, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
stx %o2, [%g2]
stx %o3, [%g3]
retl
nop
.size sun4v_ldc_tx_get_state, .-sun4v_ldc_tx_get_state
/* %o0: channel
* %o1: tail_off
*
* returns %o0: status
*/
.globl sun4v_ldc_tx_set_qtail
.type sun4v_ldc_tx_set_qtail,#function
sun4v_ldc_tx_set_qtail:
mov HV_FAST_LDC_TX_SET_QTAIL, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_tx_set_qtail, .-sun4v_ldc_tx_set_qtail
/* %o0: channel
* %o1: ra
* %o2: num_entries
*
* returns %o0: status
*/
.globl sun4v_ldc_rx_qconf
.type sun4v_ldc_rx_qconf,#function
sun4v_ldc_rx_qconf:
mov HV_FAST_LDC_RX_QCONF, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_rx_qconf, .-sun4v_ldc_rx_qconf
/* %o0: channel
* %o1: pointer to unsigned long ra
* %o2: pointer to unsigned long num_entries
*
* returns %o0: status
*/
.globl sun4v_ldc_rx_qinfo
.type sun4v_ldc_rx_qinfo,#function
sun4v_ldc_rx_qinfo:
mov %o1, %g1
mov %o2, %g2
mov HV_FAST_LDC_RX_QINFO, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
stx %o2, [%g2]
retl
nop
.size sun4v_ldc_rx_qinfo, .-sun4v_ldc_rx_qinfo
/* %o0: channel
* %o1: pointer to unsigned long head_off
* %o2: pointer to unsigned long tail_off
* %o2: pointer to unsigned long chan_state
*
* returns %o0: status
*/
.globl sun4v_ldc_rx_get_state
.type sun4v_ldc_rx_get_state,#function
sun4v_ldc_rx_get_state:
mov %o1, %g1
mov %o2, %g2
mov %o3, %g3
mov HV_FAST_LDC_RX_GET_STATE, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
stx %o2, [%g2]
stx %o3, [%g3]
retl
nop
.size sun4v_ldc_rx_get_state, .-sun4v_ldc_rx_get_state
/* %o0: channel
* %o1: head_off
*
* returns %o0: status
*/
.globl sun4v_ldc_rx_set_qhead
.type sun4v_ldc_rx_set_qhead,#function
sun4v_ldc_rx_set_qhead:
mov HV_FAST_LDC_RX_SET_QHEAD, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_rx_set_qhead, .-sun4v_ldc_rx_set_qhead
/* %o0: channel
* %o1: ra
* %o2: num_entries
*
* returns %o0: status
*/
.globl sun4v_ldc_set_map_table
.type sun4v_ldc_set_map_table,#function
sun4v_ldc_set_map_table:
mov HV_FAST_LDC_SET_MAP_TABLE, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_set_map_table, .-sun4v_ldc_set_map_table
/* %o0: channel
* %o1: pointer to unsigned long ra
* %o2: pointer to unsigned long num_entries
*
* returns %o0: status
*/
.globl sun4v_ldc_get_map_table
.type sun4v_ldc_get_map_table,#function
sun4v_ldc_get_map_table:
mov %o1, %g1
mov %o2, %g2
mov HV_FAST_LDC_GET_MAP_TABLE, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
stx %o2, [%g2]
retl
nop
.size sun4v_ldc_get_map_table, .-sun4v_ldc_get_map_table
/* %o0: channel
* %o1: dir_code
* %o2: tgt_raddr
* %o3: lcl_raddr
* %o4: len
* %o5: pointer to unsigned long actual_len
*
* returns %o0: status
*/
.globl sun4v_ldc_copy
.type sun4v_ldc_copy,#function
sun4v_ldc_copy:
mov %o5, %g1
mov HV_FAST_LDC_COPY, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
retl
nop
.size sun4v_ldc_copy, .-sun4v_ldc_copy
/* %o0: channel
* %o1: cookie
* %o2: pointer to unsigned long ra
* %o3: pointer to unsigned long perm
*
* returns %o0: status
*/
.globl sun4v_ldc_mapin
.type sun4v_ldc_mapin,#function
sun4v_ldc_mapin:
mov %o2, %g1
mov %o3, %g2
mov HV_FAST_LDC_MAPIN, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
stx %o2, [%g2]
retl
nop
.size sun4v_ldc_mapin, .-sun4v_ldc_mapin
/* %o0: ra
*
* returns %o0: status
*/
.globl sun4v_ldc_unmap
.type sun4v_ldc_unmap,#function
sun4v_ldc_unmap:
mov HV_FAST_LDC_UNMAP, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_unmap, .-sun4v_ldc_unmap
[SPARC64]: Fix args to sun4v_ldc_revoke(). First argument is LDC channel ID, then mapping cookie, then the MTE revoke cookie. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-13 08:56:52 +02:00
/* %o0: channel
* %o1: cookie
* %o2: mte_cookie
[SPARC64]: Fill holes in hypervisor APIs and fix KTSB registry. Several interfaces were missing and others misnumbered or improperly documented. Also, make sure to check the return value when registering the kernel TSBs with the hypervisor. This helped to find the 4MB kernel TSB alignment bug fixed in a previous changeset. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-29 11:22:14 +02:00
*
* returns %o0: status
*/
.globl sun4v_ldc_revoke
.type sun4v_ldc_revoke,#function
sun4v_ldc_revoke:
mov HV_FAST_LDC_REVOKE, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ldc_revoke, .-sun4v_ldc_revoke
/* %o0: device handle
* %o1: device INO
* %o2: pointer to unsigned long cookie
*
* returns %o0: status
*/
.globl sun4v_vintr_get_cookie
.type sun4v_vintr_get_cookie,#function
sun4v_vintr_get_cookie:
mov %o2, %g1
mov HV_FAST_VINTR_GET_COOKIE, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
retl
nop
.size sun4v_vintr_get_cookie, .-sun4v_vintr_get_cookie
/* %o0: device handle
* %o1: device INO
* %o2: cookie
*
* returns %o0: status
*/
.globl sun4v_vintr_set_cookie
.type sun4v_vintr_set_cookie,#function
sun4v_vintr_set_cookie:
mov HV_FAST_VINTR_SET_COOKIE, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_vintr_set_cookie, .-sun4v_vintr_set_cookie
/* %o0: device handle
* %o1: device INO
* %o2: pointer to unsigned long valid_state
*
* returns %o0: status
*/
.globl sun4v_vintr_get_valid
.type sun4v_vintr_get_valid,#function
sun4v_vintr_get_valid:
mov %o2, %g1
mov HV_FAST_VINTR_GET_VALID, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
retl
nop
.size sun4v_vintr_get_valid, .-sun4v_vintr_get_valid
/* %o0: device handle
* %o1: device INO
* %o2: valid_state
*
* returns %o0: status
*/
.globl sun4v_vintr_set_valid
.type sun4v_vintr_set_valid,#function
sun4v_vintr_set_valid:
mov HV_FAST_VINTR_SET_VALID, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_vintr_set_valid, .-sun4v_vintr_set_valid
/* %o0: device handle
* %o1: device INO
* %o2: pointer to unsigned long state
*
* returns %o0: status
*/
.globl sun4v_vintr_get_state
.type sun4v_vintr_get_state,#function
sun4v_vintr_get_state:
mov %o2, %g1
mov HV_FAST_VINTR_GET_STATE, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
retl
nop
.size sun4v_vintr_get_state, .-sun4v_vintr_get_state
/* %o0: device handle
* %o1: device INO
* %o2: state
*
* returns %o0: status
*/
.globl sun4v_vintr_set_state
.type sun4v_vintr_set_state,#function
sun4v_vintr_set_state:
mov HV_FAST_VINTR_SET_STATE, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_vintr_set_state, .-sun4v_vintr_set_state
/* %o0: device handle
* %o1: device INO
* %o2: pointer to unsigned long cpuid
*
* returns %o0: status
*/
.globl sun4v_vintr_get_target
.type sun4v_vintr_get_target,#function
sun4v_vintr_get_target:
mov %o2, %g1
mov HV_FAST_VINTR_GET_TARGET, %o5
ta HV_FAST_TRAP
stx %o1, [%g1]
retl
nop
.size sun4v_vintr_get_target, .-sun4v_vintr_get_target
/* %o0: device handle
* %o1: device INO
* %o2: cpuid
*
* returns %o0: status
*/
.globl sun4v_vintr_set_target
.type sun4v_vintr_set_target,#function
sun4v_vintr_set_target:
mov HV_FAST_VINTR_SET_TARGET, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_vintr_set_target, .-sun4v_vintr_set_target
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
/* %o0: NCS sub-function
* %o1: sub-function arg real-address
* %o2: sub-function arg size
*
* returns %o0: status
*/
.globl sun4v_ncs_request
.type sun4v_ncs_request,#function
sun4v_ncs_request:
mov HV_FAST_NCS_REQUEST, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_ncs_request, .-sun4v_ncs_request
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.globl sun4v_svc_send
.type sun4v_svc_send,#function
sun4v_svc_send:
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
save %sp, -192, %sp
mov %i0, %o0
mov %i1, %o1
mov %i2, %o2
mov HV_FAST_SVC_SEND, %o5
ta HV_FAST_TRAP
stx %o1, [%i3]
ret
restore
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.size sun4v_svc_send, .-sun4v_svc_send
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.globl sun4v_svc_recv
.type sun4v_svc_recv,#function
sun4v_svc_recv:
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
save %sp, -192, %sp
mov %i0, %o0
mov %i1, %o1
mov %i2, %o2
mov HV_FAST_SVC_RECV, %o5
ta HV_FAST_TRAP
stx %o1, [%i3]
ret
restore
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.size sun4v_svc_recv, .-sun4v_svc_recv
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.globl sun4v_svc_getstatus
.type sun4v_svc_getstatus,#function
sun4v_svc_getstatus:
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
mov HV_FAST_SVC_GETSTATUS, %o5
mov %o1, %o4
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.size sun4v_svc_getstatus, .-sun4v_svc_getstatus
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.globl sun4v_svc_setstatus
.type sun4v_svc_setstatus,#function
sun4v_svc_setstatus:
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
mov HV_FAST_SVC_SETSTATUS, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.size sun4v_svc_setstatus, .-sun4v_svc_setstatus
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.globl sun4v_svc_clrstatus
.type sun4v_svc_clrstatus,#function
sun4v_svc_clrstatus:
[SPARC64]: Add missing NCS and SVC hypervisor interfaces. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-05-31 04:01:47 +02:00
mov HV_FAST_SVC_CLRSTATUS, %o5
ta HV_FAST_TRAP
retl
nop
[SPARC64]: Fix service channel hypervisor function names. sed 's/scv/svc/' Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 04:07:32 +02:00
.size sun4v_svc_clrstatus, .-sun4v_svc_clrstatus
[SPARC64]: Provide mmu statistics via sysfs. If the system supports hypervisor based statistics, allow them to be fetched, enabled, and disabled via sysfs. Enable and disable via the boolean: /sys/devices/systems/cpu/cpuN/mmustat_enable Statistic values are provided under: /sys/devices/systems/cpu/cpuN/mmu_status/ Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-04 08:38:09 +02:00
.globl sun4v_mmustat_conf
.type sun4v_mmustat_conf,#function
sun4v_mmustat_conf:
mov %o1, %o4
mov HV_FAST_MMUSTAT_CONF, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
.size sun4v_mmustat_conf, .-sun4v_mmustat_conf
.globl sun4v_mmustat_info
.type sun4v_mmustat_info,#function
sun4v_mmustat_info:
mov %o0, %o4
mov HV_FAST_MMUSTAT_INFO, %o5
ta HV_FAST_TRAP
stx %o1, [%o4]
retl
nop
.size sun4v_mmustat_info, .-sun4v_mmustat_info
[SPARC64]: Fix two kernel linear mapping setup bugs. This was caught and identified by Greg Onufer. Since we setup the 256M/4M bitmap table after taking over the trap table, it's possible for some 4M mapping to get loaded in the TLB beforhand which later will be 256M mappings. This can cause illegal TLB multiple-match conditions. Fix this by setting up the bitmap before we take over the trap table. Next, __flush_tlb_all() was not doing anything on hypervisor platforms. Fix by adding sun4v_mmu_demap_all() and calling it. Signed-off-by: David S. Miller <davem@davemloft.net>
2007-12-13 15:13:38 +01:00
.globl sun4v_mmu_demap_all
.type sun4v_mmu_demap_all,#function
sun4v_mmu_demap_all:
clr %o0
clr %o1
mov HV_MMU_ALL, %o2
mov HV_FAST_MMU_DEMAP_ALL, %o5
ta HV_FAST_TRAP
retl
nop
.size sun4v_mmu_demap_all, .-sun4v_mmu_demap_all
Reference in a new issue Copy permalink