linux/arch/powerpc/kernel/entry_64.S

1272 lines
30 KiB
ArmAsm
Raw Normal View History

/*
* PowerPC version
* Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
* Rewritten by Cort Dougan (cort@cs.nmt.edu) for PReP
* Copyright (C) 1996 Cort Dougan <cort@cs.nmt.edu>
* Adapted for Power Macintosh by Paul Mackerras.
* Low-level exception handlers and MMU support
* rewritten by Paul Mackerras.
* Copyright (C) 1996 Paul Mackerras.
* MPC8xx modifications Copyright (C) 1997 Dan Malek (dmalek@jlc.net).
*
* This file contains the system call entry code, context switch
* code, and exception/interrupt return code for PowerPC.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#include <linux/errno.h>
#include <linux/err.h>
#include <asm/unistd.h>
#include <asm/processor.h>
#include <asm/page.h>
#include <asm/mmu.h>
#include <asm/thread_info.h>
#include <asm/ppc_asm.h>
#include <asm/asm-offsets.h>
#include <asm/cputable.h>
#include <asm/firmware.h>
#include <asm/bug.h>
#include <asm/ptrace.h>
#include <asm/irqflags.h>
#include <asm/ftrace.h>
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
#include <asm/hw_irq.h>
#include <asm/context_tracking.h>
#include <asm/tm.h>
/*
* System calls.
*/
.section ".toc","aw"
SYS_CALL_TABLE:
.tc sys_call_table[TC],sys_call_table
/* This value is used to mark exception frames on the stack. */
exception_marker:
.tc ID_EXC_MARKER[TC],STACK_FRAME_REGS_MARKER
.section ".text"
.align 7
.globl system_call_common
system_call_common:
#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
BEGIN_FTR_SECTION
extrdi. r10, r12, 1, (63-MSR_TS_T_LG) /* transaction active? */
bne tabort_syscall
END_FTR_SECTION_IFSET(CPU_FTR_TM)
#endif
andi. r10,r12,MSR_PR
mr r10,r1
addi r1,r1,-INT_FRAME_SIZE
beq- 1f
ld r1,PACAKSAVE(r13)
1: std r10,0(r1)
std r11,_NIP(r1)
std r12,_MSR(r1)
std r0,GPR0(r1)
std r10,GPR1(r1)
beq 2f /* if from kernel mode */
powerpc: Implement accurate task and CPU time accounting This implements accurate task and cpu time accounting for 64-bit powerpc kernels. Instead of accounting a whole jiffy of time to a task on a timer interrupt because that task happened to be running at the time, we now account time in units of timebase ticks according to the actual time spent by the task in user mode and kernel mode. We also count the time spent processing hardware and software interrupts accurately. This is conditional on CONFIG_VIRT_CPU_ACCOUNTING. If that is not set, we do tick-based approximate accounting as before. To get this accurate information, we read either the PURR (processor utilization of resources register) on POWER5 machines, or the timebase on other machines on * each entry to the kernel from usermode * each exit to usermode * transitions between process context, hard irq context and soft irq context in kernel mode * context switches. On POWER5 systems with shared-processor logical partitioning we also read both the PURR and the timebase at each timer interrupt and context switch in order to determine how much time has been taken by the hypervisor to run other partitions ("steal" time). Unfortunately, since we need values of the PURR on both threads at the same time to accurately calculate the steal time, and since we can only calculate steal time on a per-core basis, the apportioning of the steal time between idle time (time which we ceded to the hypervisor in the idle loop) and actual stolen time is somewhat approximate at the moment. This is all based quite heavily on what s390 does, and it uses the generic interfaces that were added by the s390 developers, i.e. account_system_time(), account_user_time(), etc. This patch doesn't add any new interfaces between the kernel and userspace, and doesn't change the units in which time is reported to userspace by things such as /proc/stat, /proc/<pid>/stat, getrusage(), times(), etc. Internally the various task and cpu times are stored in timebase units, but they are converted to USER_HZ units (1/100th of a second) when reported to userspace. Some precision is therefore lost but there should not be any accumulating error, since the internal accumulation is at full precision. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-02-24 07:06:59 +08:00
ACCOUNT_CPU_USER_ENTRY(r10, r11)
2: std r2,GPR2(r1)
std r3,GPR3(r1)
mfcr r2
std r4,GPR4(r1)
std r5,GPR5(r1)
std r6,GPR6(r1)
std r7,GPR7(r1)
std r8,GPR8(r1)
li r11,0
std r11,GPR9(r1)
std r11,GPR10(r1)
std r11,GPR11(r1)
std r11,GPR12(r1)
std r11,_XER(r1)
std r11,_CTR(r1)
std r9,GPR13(r1)
mflr r10
/*
* This clears CR0.SO (bit 28), which is the error indication on
* return from this system call.
*/
rldimi r2,r11,28,(63-28)
li r11,0xc01
std r10,_LINK(r1)
std r11,_TRAP(r1)
std r3,ORIG_GPR3(r1)
std r2,_CCR(r1)
ld r2,PACATOC(r13)
addi r9,r1,STACK_FRAME_OVERHEAD
ld r11,exception_marker@toc(r2)
std r11,-16(r9) /* "regshere" marker */
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 13:56:04 +08:00
#if defined(CONFIG_VIRT_CPU_ACCOUNTING_NATIVE) && defined(CONFIG_PPC_SPLPAR)
powerpc: Account time using timebase rather than PURR Currently, when CONFIG_VIRT_CPU_ACCOUNTING is enabled, we use the PURR register for measuring the user and system time used by processes, as well as other related times such as hardirq and softirq times. This turns out to be quite confusing for users because it means that a program will often be measured as taking less time when run on a multi-threaded processor (SMT2 or SMT4 mode) than it does when run on a single-threaded processor (ST mode), even though the program takes longer to finish. The discrepancy is accounted for as stolen time, which is also confusing, particularly when there are no other partitions running. This changes the accounting to use the timebase instead, meaning that the reported user and system times are the actual number of real-time seconds that the program was executing on the processor thread, regardless of which SMT mode the processor is in. Thus a program will generally show greater user and system times when run on a multi-threaded processor than on a single-threaded processor. On pSeries systems on POWER5 or later processors, we measure the stolen time (time when this partition wasn't running) using the hypervisor dispatch trace log. We check for new entries in the log on every entry from user mode and on every transition from kernel process context to soft or hard IRQ context (i.e. when account_system_vtime() gets called). So that we can correctly distinguish time stolen from user time and time stolen from system time, without having to check the log on every exit to user mode, we store separate timestamps for exit to user mode and entry from user mode. On systems that have a SPURR (POWER6 and POWER7), we read the SPURR in account_system_vtime() (as before), and then apportion the SPURR ticks since the last time we read it between scaled user time and scaled system time according to the relative proportions of user time and system time over the same interval. This avoids having to read the SPURR on every kernel entry and exit. On systems that have PURR but not SPURR (i.e., POWER5), we do the same using the PURR rather than the SPURR. This disables the DTL user interface in /sys/debug/kernel/powerpc/dtl for now since it conflicts with the use of the dispatch trace log by the time accounting code. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-27 03:56:43 +08:00
BEGIN_FW_FTR_SECTION
beq 33f
/* if from user, see if there are any DTL entries to process */
ld r10,PACALPPACAPTR(r13) /* get ptr to VPA */
ld r11,PACA_DTL_RIDX(r13) /* get log read index */
addi r10,r10,LPPACA_DTLIDX
LDX_BE r10,0,r10 /* get log write index */
powerpc: Account time using timebase rather than PURR Currently, when CONFIG_VIRT_CPU_ACCOUNTING is enabled, we use the PURR register for measuring the user and system time used by processes, as well as other related times such as hardirq and softirq times. This turns out to be quite confusing for users because it means that a program will often be measured as taking less time when run on a multi-threaded processor (SMT2 or SMT4 mode) than it does when run on a single-threaded processor (ST mode), even though the program takes longer to finish. The discrepancy is accounted for as stolen time, which is also confusing, particularly when there are no other partitions running. This changes the accounting to use the timebase instead, meaning that the reported user and system times are the actual number of real-time seconds that the program was executing on the processor thread, regardless of which SMT mode the processor is in. Thus a program will generally show greater user and system times when run on a multi-threaded processor than on a single-threaded processor. On pSeries systems on POWER5 or later processors, we measure the stolen time (time when this partition wasn't running) using the hypervisor dispatch trace log. We check for new entries in the log on every entry from user mode and on every transition from kernel process context to soft or hard IRQ context (i.e. when account_system_vtime() gets called). So that we can correctly distinguish time stolen from user time and time stolen from system time, without having to check the log on every exit to user mode, we store separate timestamps for exit to user mode and entry from user mode. On systems that have a SPURR (POWER6 and POWER7), we read the SPURR in account_system_vtime() (as before), and then apportion the SPURR ticks since the last time we read it between scaled user time and scaled system time according to the relative proportions of user time and system time over the same interval. This avoids having to read the SPURR on every kernel entry and exit. On systems that have PURR but not SPURR (i.e., POWER5), we do the same using the PURR rather than the SPURR. This disables the DTL user interface in /sys/debug/kernel/powerpc/dtl for now since it conflicts with the use of the dispatch trace log by the time accounting code. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-27 03:56:43 +08:00
cmpd cr1,r11,r10
beq+ cr1,33f
bl accumulate_stolen_time
powerpc: Account time using timebase rather than PURR Currently, when CONFIG_VIRT_CPU_ACCOUNTING is enabled, we use the PURR register for measuring the user and system time used by processes, as well as other related times such as hardirq and softirq times. This turns out to be quite confusing for users because it means that a program will often be measured as taking less time when run on a multi-threaded processor (SMT2 or SMT4 mode) than it does when run on a single-threaded processor (ST mode), even though the program takes longer to finish. The discrepancy is accounted for as stolen time, which is also confusing, particularly when there are no other partitions running. This changes the accounting to use the timebase instead, meaning that the reported user and system times are the actual number of real-time seconds that the program was executing on the processor thread, regardless of which SMT mode the processor is in. Thus a program will generally show greater user and system times when run on a multi-threaded processor than on a single-threaded processor. On pSeries systems on POWER5 or later processors, we measure the stolen time (time when this partition wasn't running) using the hypervisor dispatch trace log. We check for new entries in the log on every entry from user mode and on every transition from kernel process context to soft or hard IRQ context (i.e. when account_system_vtime() gets called). So that we can correctly distinguish time stolen from user time and time stolen from system time, without having to check the log on every exit to user mode, we store separate timestamps for exit to user mode and entry from user mode. On systems that have a SPURR (POWER6 and POWER7), we read the SPURR in account_system_vtime() (as before), and then apportion the SPURR ticks since the last time we read it between scaled user time and scaled system time according to the relative proportions of user time and system time over the same interval. This avoids having to read the SPURR on every kernel entry and exit. On systems that have PURR but not SPURR (i.e., POWER5), we do the same using the PURR rather than the SPURR. This disables the DTL user interface in /sys/debug/kernel/powerpc/dtl for now since it conflicts with the use of the dispatch trace log by the time accounting code. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-27 03:56:43 +08:00
REST_GPR(0,r1)
REST_4GPRS(3,r1)
REST_2GPRS(7,r1)
addi r9,r1,STACK_FRAME_OVERHEAD
33:
END_FW_FTR_SECTION_IFSET(FW_FEATURE_SPLPAR)
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 13:56:04 +08:00
#endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE && CONFIG_PPC_SPLPAR */
powerpc: Account time using timebase rather than PURR Currently, when CONFIG_VIRT_CPU_ACCOUNTING is enabled, we use the PURR register for measuring the user and system time used by processes, as well as other related times such as hardirq and softirq times. This turns out to be quite confusing for users because it means that a program will often be measured as taking less time when run on a multi-threaded processor (SMT2 or SMT4 mode) than it does when run on a single-threaded processor (ST mode), even though the program takes longer to finish. The discrepancy is accounted for as stolen time, which is also confusing, particularly when there are no other partitions running. This changes the accounting to use the timebase instead, meaning that the reported user and system times are the actual number of real-time seconds that the program was executing on the processor thread, regardless of which SMT mode the processor is in. Thus a program will generally show greater user and system times when run on a multi-threaded processor than on a single-threaded processor. On pSeries systems on POWER5 or later processors, we measure the stolen time (time when this partition wasn't running) using the hypervisor dispatch trace log. We check for new entries in the log on every entry from user mode and on every transition from kernel process context to soft or hard IRQ context (i.e. when account_system_vtime() gets called). So that we can correctly distinguish time stolen from user time and time stolen from system time, without having to check the log on every exit to user mode, we store separate timestamps for exit to user mode and entry from user mode. On systems that have a SPURR (POWER6 and POWER7), we read the SPURR in account_system_vtime() (as before), and then apportion the SPURR ticks since the last time we read it between scaled user time and scaled system time according to the relative proportions of user time and system time over the same interval. This avoids having to read the SPURR on every kernel entry and exit. On systems that have PURR but not SPURR (i.e., POWER5), we do the same using the PURR rather than the SPURR. This disables the DTL user interface in /sys/debug/kernel/powerpc/dtl for now since it conflicts with the use of the dispatch trace log by the time accounting code. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-27 03:56:43 +08:00
/*
* A syscall should always be called with interrupts enabled
* so we just unconditionally hard-enable here. When some kind
* of irq tracing is used, we additionally check that condition
* is correct
*/
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_BUG)
lbz r10,PACASOFTIRQEN(r13)
xori r10,r10,1
1: tdnei r10,0
EMIT_BUG_ENTRY 1b,__FILE__,__LINE__,BUGFLAG_WARNING
#endif
#ifdef CONFIG_PPC_BOOK3E
wrteei 1
#else
ld r11,PACAKMSR(r13)
ori r11,r11,MSR_EE
mtmsrd r11,1
#endif /* CONFIG_PPC_BOOK3E */
/* We do need to set SOFTE in the stack frame or the return
* from interrupt will be painful
*/
li r10,1
std r10,SOFTE(r1)
CURRENT_THREAD_INFO(r11, r1)
ld r10,TI_FLAGS(r11)
andi. r11,r10,_TIF_SYSCALL_DOTRACE
bne syscall_dotrace /* does not return */
cmpldi 0,r0,NR_syscalls
bge- syscall_enosys
system_call: /* label this so stack traces look sane */
/*
* Need to vector to 32 Bit or default sys_call_table here,
* based on caller's run-mode / personality.
*/
ld r11,SYS_CALL_TABLE@toc(2)
andi. r10,r10,_TIF_32BIT
beq 15f
addi r11,r11,8 /* use 32-bit syscall entries */
clrldi r3,r3,32
clrldi r4,r4,32
clrldi r5,r5,32
clrldi r6,r6,32
clrldi r7,r7,32
clrldi r8,r8,32
15:
slwi r0,r0,4
ldx r12,r11,r0 /* Fetch system call handler [ptr] */
mtctr r12
bctrl /* Call handler */
.Lsyscall_exit:
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
std r3,RESULT(r1)
CURRENT_THREAD_INFO(r12, r1)
ld r8,_MSR(r1)
#ifdef CONFIG_PPC_BOOK3S
/* No MSR:RI on BookE */
andi. r10,r8,MSR_RI
beq- unrecov_restore
#endif
/*
* Disable interrupts so current_thread_info()->flags can't change,
* and so that we don't get interrupted after loading SRR0/1.
*/
#ifdef CONFIG_PPC_BOOK3E
wrteei 0
#else
ld r10,PACAKMSR(r13)
/*
* For performance reasons we clear RI the same time that we
* clear EE. We only need to clear RI just before we restore r13
* below, but batching it with EE saves us one expensive mtmsrd call.
* We have to be careful to restore RI if we branch anywhere from
* here (eg syscall_exit_work).
*/
li r9,MSR_RI
andc r11,r10,r9
mtmsrd r11,1
#endif /* CONFIG_PPC_BOOK3E */
ld r9,TI_FLAGS(r12)
li r11,-MAX_ERRNO
andi. r0,r9,(_TIF_SYSCALL_DOTRACE|_TIF_SINGLESTEP|_TIF_USER_WORK_MASK|_TIF_PERSYSCALL_MASK)
bne- syscall_exit_work
powerpc: Restore FPU/VEC/VSX if previously used Currently the FPU, VEC and VSX facilities are lazily loaded. This is not a problem unless a process is using these facilities. Modern versions of GCC are very good at automatically vectorising code, new and modernised workloads make use of floating point and vector facilities, even the kernel makes use of vectorised memcpy. All this combined greatly increases the cost of a syscall since the kernel uses the facilities sometimes even in syscall fast-path making it increasingly common for a thread to take an *_unavailable exception soon after a syscall, not to mention potentially taking all three. The obvious overcompensation to this problem is to simply always load all the facilities on every exit to userspace. Loading up all FPU, VEC and VSX registers every time can be expensive and if a workload does avoid using them, it should not be forced to incur this penalty. An 8bit counter is used to detect if the registers have been used in the past and the registers are always loaded until the value wraps to back to zero. Several versions of the assembly in entry_64.S were tested: 1. Always calling C. 2. Performing a common case check and then calling C. 3. A complex check in asm. After some benchmarking it was determined that avoiding C in the common case is a performance benefit (option 2). The full check in asm (option 3) greatly complicated that codepath for a negligible performance gain and the trade-off was deemed not worth it. Signed-off-by: Cyril Bur <cyrilbur@gmail.com> [mpe: Move load_vec in the struct to fill an existing hole, reword change log] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> fixup
2016-02-29 14:53:47 +08:00
andi. r0,r8,MSR_FP
beq 2f
#ifdef CONFIG_ALTIVEC
andis. r0,r8,MSR_VEC@h
bne 3f
#endif
2: addi r3,r1,STACK_FRAME_OVERHEAD
bl restore_math
ld r8,_MSR(r1)
ld r3,RESULT(r1)
li r11,-MAX_ERRNO
3: cmpld r3,r11
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
ld r5,_CCR(r1)
bge- syscall_error
.Lsyscall_error_cont:
ld r7,_NIP(r1)
powerpc: Feature nop out reservation clear when stcx checks address The POWER architecture does not require stcx to check that it is operating on the same address as the larx. This means it is possible for an an exception handler to execute a larx, get a reservation, decide not to do the stcx and then return back with an active reservation. If the interrupted code was in the middle of a larx/stcx sequence the stcx could incorrectly succeed. All recent POWER CPUs check the address before letting the stcx succeed so we can create a CPU feature and nop it out. As Ben suggested, we can only do this in our syscall path because there is a remote possibility some kernel code gets interrupted by an exception that ends up operating on the same cacheline. Thanks to Paul Mackerras and Derek Williams for the idea. To test this I used a very simple null syscall (actually getppid) testcase at http://ozlabs.org/~anton/junkcode/null_syscall.c I tested against 2.6.35-git10 with the following changes against the pseries_defconfig: CONFIG_VIRT_CPU_ACCOUNTING=n CONFIG_AUDIT=n CONFIG_PPC_4K_PAGES=n CONFIG_PPC_64K_PAGES=y CONFIG_FORCE_MAX_ZONEORDER=9 CONFIG_PPC_SUBPAGE_PROT=n CONFIG_FUNCTION_TRACER=n CONFIG_FUNCTION_GRAPH_TRACER=n CONFIG_IRQSOFF_TRACER=n CONFIG_STACK_TRACER=n to remove the overhead of virtual CPU accounting, syscall auditing and the ftrace mcount tracers. 64kB pages were enabled to minimise TLB misses. POWER6: +8.2% POWER7: +7.0% Another suggestion was to use a larx to something in the L1 instead of a stcx. This was almost as fast as removing the larx on POWER6, but only 3.5% faster on POWER7. We can use this to speed up the reservation clear in our exception exit code. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-11 09:40:27 +08:00
BEGIN_FTR_SECTION
stdcx. r0,0,r1 /* to clear the reservation */
powerpc: Feature nop out reservation clear when stcx checks address The POWER architecture does not require stcx to check that it is operating on the same address as the larx. This means it is possible for an an exception handler to execute a larx, get a reservation, decide not to do the stcx and then return back with an active reservation. If the interrupted code was in the middle of a larx/stcx sequence the stcx could incorrectly succeed. All recent POWER CPUs check the address before letting the stcx succeed so we can create a CPU feature and nop it out. As Ben suggested, we can only do this in our syscall path because there is a remote possibility some kernel code gets interrupted by an exception that ends up operating on the same cacheline. Thanks to Paul Mackerras and Derek Williams for the idea. To test this I used a very simple null syscall (actually getppid) testcase at http://ozlabs.org/~anton/junkcode/null_syscall.c I tested against 2.6.35-git10 with the following changes against the pseries_defconfig: CONFIG_VIRT_CPU_ACCOUNTING=n CONFIG_AUDIT=n CONFIG_PPC_4K_PAGES=n CONFIG_PPC_64K_PAGES=y CONFIG_FORCE_MAX_ZONEORDER=9 CONFIG_PPC_SUBPAGE_PROT=n CONFIG_FUNCTION_TRACER=n CONFIG_FUNCTION_GRAPH_TRACER=n CONFIG_IRQSOFF_TRACER=n CONFIG_STACK_TRACER=n to remove the overhead of virtual CPU accounting, syscall auditing and the ftrace mcount tracers. 64kB pages were enabled to minimise TLB misses. POWER6: +8.2% POWER7: +7.0% Another suggestion was to use a larx to something in the L1 instead of a stcx. This was almost as fast as removing the larx on POWER6, but only 3.5% faster on POWER7. We can use this to speed up the reservation clear in our exception exit code. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-11 09:40:27 +08:00
END_FTR_SECTION_IFCLR(CPU_FTR_STCX_CHECKS_ADDRESS)
andi. r6,r8,MSR_PR
ld r4,_LINK(r1)
powerpc: Implement accurate task and CPU time accounting This implements accurate task and cpu time accounting for 64-bit powerpc kernels. Instead of accounting a whole jiffy of time to a task on a timer interrupt because that task happened to be running at the time, we now account time in units of timebase ticks according to the actual time spent by the task in user mode and kernel mode. We also count the time spent processing hardware and software interrupts accurately. This is conditional on CONFIG_VIRT_CPU_ACCOUNTING. If that is not set, we do tick-based approximate accounting as before. To get this accurate information, we read either the PURR (processor utilization of resources register) on POWER5 machines, or the timebase on other machines on * each entry to the kernel from usermode * each exit to usermode * transitions between process context, hard irq context and soft irq context in kernel mode * context switches. On POWER5 systems with shared-processor logical partitioning we also read both the PURR and the timebase at each timer interrupt and context switch in order to determine how much time has been taken by the hypervisor to run other partitions ("steal" time). Unfortunately, since we need values of the PURR on both threads at the same time to accurately calculate the steal time, and since we can only calculate steal time on a per-core basis, the apportioning of the steal time between idle time (time which we ceded to the hypervisor in the idle loop) and actual stolen time is somewhat approximate at the moment. This is all based quite heavily on what s390 does, and it uses the generic interfaces that were added by the s390 developers, i.e. account_system_time(), account_user_time(), etc. This patch doesn't add any new interfaces between the kernel and userspace, and doesn't change the units in which time is reported to userspace by things such as /proc/stat, /proc/<pid>/stat, getrusage(), times(), etc. Internally the various task and cpu times are stored in timebase units, but they are converted to USER_HZ units (1/100th of a second) when reported to userspace. Some precision is therefore lost but there should not be any accumulating error, since the internal accumulation is at full precision. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-02-24 07:06:59 +08:00
beq- 1f
ACCOUNT_CPU_USER_EXIT(r11, r12)
BEGIN_FTR_SECTION
HMT_MEDIUM_LOW
END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
powerpc: Implement accurate task and CPU time accounting This implements accurate task and cpu time accounting for 64-bit powerpc kernels. Instead of accounting a whole jiffy of time to a task on a timer interrupt because that task happened to be running at the time, we now account time in units of timebase ticks according to the actual time spent by the task in user mode and kernel mode. We also count the time spent processing hardware and software interrupts accurately. This is conditional on CONFIG_VIRT_CPU_ACCOUNTING. If that is not set, we do tick-based approximate accounting as before. To get this accurate information, we read either the PURR (processor utilization of resources register) on POWER5 machines, or the timebase on other machines on * each entry to the kernel from usermode * each exit to usermode * transitions between process context, hard irq context and soft irq context in kernel mode * context switches. On POWER5 systems with shared-processor logical partitioning we also read both the PURR and the timebase at each timer interrupt and context switch in order to determine how much time has been taken by the hypervisor to run other partitions ("steal" time). Unfortunately, since we need values of the PURR on both threads at the same time to accurately calculate the steal time, and since we can only calculate steal time on a per-core basis, the apportioning of the steal time between idle time (time which we ceded to the hypervisor in the idle loop) and actual stolen time is somewhat approximate at the moment. This is all based quite heavily on what s390 does, and it uses the generic interfaces that were added by the s390 developers, i.e. account_system_time(), account_user_time(), etc. This patch doesn't add any new interfaces between the kernel and userspace, and doesn't change the units in which time is reported to userspace by things such as /proc/stat, /proc/<pid>/stat, getrusage(), times(), etc. Internally the various task and cpu times are stored in timebase units, but they are converted to USER_HZ units (1/100th of a second) when reported to userspace. Some precision is therefore lost but there should not be any accumulating error, since the internal accumulation is at full precision. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-02-24 07:06:59 +08:00
ld r13,GPR13(r1) /* only restore r13 if returning to usermode */
1: ld r2,GPR2(r1)
ld r1,GPR1(r1)
mtlr r4
mtcr r5
mtspr SPRN_SRR0,r7
mtspr SPRN_SRR1,r8
RFI
b . /* prevent speculative execution */
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
syscall_error:
oris r5,r5,0x1000 /* Set SO bit in CR */
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
neg r3,r3
std r5,_CCR(r1)
b .Lsyscall_error_cont
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
/* Traced system call support */
syscall_dotrace:
bl save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
bl do_syscall_trace_enter
/*
* We use the return value of do_syscall_trace_enter() as the syscall
* number. If the syscall was rejected for any reason do_syscall_trace_enter()
* returns an invalid syscall number and the test below against
* NR_syscalls will fail.
*/
mr r0,r3
/* Restore argument registers just clobbered and/or possibly changed. */
ld r3,GPR3(r1)
ld r4,GPR4(r1)
ld r5,GPR5(r1)
ld r6,GPR6(r1)
ld r7,GPR7(r1)
ld r8,GPR8(r1)
/* Repopulate r9 and r10 for the system_call path */
addi r9,r1,STACK_FRAME_OVERHEAD
CURRENT_THREAD_INFO(r10, r1)
ld r10,TI_FLAGS(r10)
cmpldi r0,NR_syscalls
blt+ system_call
/* Return code is already in r3 thanks to do_syscall_trace_enter() */
b .Lsyscall_exit
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
syscall_enosys:
li r3,-ENOSYS
b .Lsyscall_exit
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
syscall_exit_work:
#ifdef CONFIG_PPC_BOOK3S
mtmsrd r10,1 /* Restore RI */
#endif
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
/* If TIF_RESTOREALL is set, don't scribble on either r3 or ccr.
If TIF_NOERROR is set, just save r3 as it is. */
andi. r0,r9,_TIF_RESTOREALL
powerpc: Fix various syscall/signal/swapcontext bugs A careful reading of the recent changes to the system call entry/exit paths revealed several problems, plus some things that could be simplified and improved: * 32-bit wasn't testing the _TIF_NOERROR bit in the syscall fast exit path, so it was only doing anything with it once it saw some other bit being set. In other words, the noerror behaviour would apply to the next system call where we had to reschedule or deliver a signal, which is not necessarily the current system call. * 32-bit wasn't doing the call to ptrace_notify in the syscall exit path when the _TIF_SINGLESTEP bit was set. * _TIF_RESTOREALL was in both _TIF_USER_WORK_MASK and _TIF_PERSYSCALL_MASK, which is odd since _TIF_RESTOREALL is only set by system calls. I took it out of _TIF_USER_WORK_MASK. * On 64-bit, _TIF_RESTOREALL wasn't causing the non-volatile registers to be restored (unless perhaps a signal was delivered or the syscall was traced or single-stepped). Thus the non-volatile registers weren't restored on exit from a signal handler. We probably got away with it mostly because signal handlers written in C wouldn't alter the non-volatile registers. * On 32-bit I simplified the code and made it more like 64-bit by making the syscall exit path jump to ret_from_except to handle preemption and signal delivery. * 32-bit was calling do_signal unnecessarily when _TIF_RESTOREALL was set - but I think because of that 32-bit was actually restoring the non-volatile registers on exit from a signal handler. * I changed the order of enabling interrupts and saving the non-volatile registers before calling do_syscall_trace_leave; now we enable interrupts first. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-03-08 10:24:22 +08:00
beq+ 0f
REST_NVGPRS(r1)
b 2f
0: cmpld r3,r11 /* r11 is -MAX_ERRNO */
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
blt+ 1f
andi. r0,r9,_TIF_NOERROR
bne- 1f
ld r5,_CCR(r1)
neg r3,r3
oris r5,r5,0x1000 /* Set SO bit in CR */
std r5,_CCR(r1)
1: std r3,GPR3(r1)
2: andi. r0,r9,(_TIF_PERSYSCALL_MASK)
beq 4f
powerpc: Fix various syscall/signal/swapcontext bugs A careful reading of the recent changes to the system call entry/exit paths revealed several problems, plus some things that could be simplified and improved: * 32-bit wasn't testing the _TIF_NOERROR bit in the syscall fast exit path, so it was only doing anything with it once it saw some other bit being set. In other words, the noerror behaviour would apply to the next system call where we had to reschedule or deliver a signal, which is not necessarily the current system call. * 32-bit wasn't doing the call to ptrace_notify in the syscall exit path when the _TIF_SINGLESTEP bit was set. * _TIF_RESTOREALL was in both _TIF_USER_WORK_MASK and _TIF_PERSYSCALL_MASK, which is odd since _TIF_RESTOREALL is only set by system calls. I took it out of _TIF_USER_WORK_MASK. * On 64-bit, _TIF_RESTOREALL wasn't causing the non-volatile registers to be restored (unless perhaps a signal was delivered or the syscall was traced or single-stepped). Thus the non-volatile registers weren't restored on exit from a signal handler. We probably got away with it mostly because signal handlers written in C wouldn't alter the non-volatile registers. * On 32-bit I simplified the code and made it more like 64-bit by making the syscall exit path jump to ret_from_except to handle preemption and signal delivery. * 32-bit was calling do_signal unnecessarily when _TIF_RESTOREALL was set - but I think because of that 32-bit was actually restoring the non-volatile registers on exit from a signal handler. * I changed the order of enabling interrupts and saving the non-volatile registers before calling do_syscall_trace_leave; now we enable interrupts first. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-03-08 10:24:22 +08:00
/* Clear per-syscall TIF flags if any are set. */
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
li r11,_TIF_PERSYSCALL_MASK
addi r12,r12,TI_FLAGS
3: ldarx r10,0,r12
andc r10,r10,r11
stdcx. r10,0,r12
bne- 3b
subi r12,r12,TI_FLAGS
powerpc: Fix various syscall/signal/swapcontext bugs A careful reading of the recent changes to the system call entry/exit paths revealed several problems, plus some things that could be simplified and improved: * 32-bit wasn't testing the _TIF_NOERROR bit in the syscall fast exit path, so it was only doing anything with it once it saw some other bit being set. In other words, the noerror behaviour would apply to the next system call where we had to reschedule or deliver a signal, which is not necessarily the current system call. * 32-bit wasn't doing the call to ptrace_notify in the syscall exit path when the _TIF_SINGLESTEP bit was set. * _TIF_RESTOREALL was in both _TIF_USER_WORK_MASK and _TIF_PERSYSCALL_MASK, which is odd since _TIF_RESTOREALL is only set by system calls. I took it out of _TIF_USER_WORK_MASK. * On 64-bit, _TIF_RESTOREALL wasn't causing the non-volatile registers to be restored (unless perhaps a signal was delivered or the syscall was traced or single-stepped). Thus the non-volatile registers weren't restored on exit from a signal handler. We probably got away with it mostly because signal handlers written in C wouldn't alter the non-volatile registers. * On 32-bit I simplified the code and made it more like 64-bit by making the syscall exit path jump to ret_from_except to handle preemption and signal delivery. * 32-bit was calling do_signal unnecessarily when _TIF_RESTOREALL was set - but I think because of that 32-bit was actually restoring the non-volatile registers on exit from a signal handler. * I changed the order of enabling interrupts and saving the non-volatile registers before calling do_syscall_trace_leave; now we enable interrupts first. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-03-08 10:24:22 +08:00
4: /* Anything else left to do? */
BEGIN_FTR_SECTION
lis r3,INIT_PPR@highest /* Set thread.ppr = 3 */
ld r10,PACACURRENT(r13)
sldi r3,r3,32 /* bits 11-13 are used for ppr */
std r3,TASKTHREADPPR(r10)
END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
andi. r0,r9,(_TIF_SYSCALL_DOTRACE|_TIF_SINGLESTEP)
beq ret_from_except_lite
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
/* Re-enable interrupts */
#ifdef CONFIG_PPC_BOOK3E
wrteei 1
#else
ld r10,PACAKMSR(r13)
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
ori r10,r10,MSR_EE
mtmsrd r10,1
#endif /* CONFIG_PPC_BOOK3E */
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
bl save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
bl do_syscall_trace_leave
b ret_from_except
#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
tabort_syscall:
/* Firstly we need to enable TM in the kernel */
mfmsr r10
li r13, 1
rldimi r10, r13, MSR_TM_LG, 63-MSR_TM_LG
mtmsrd r10, 0
/* tabort, this dooms the transaction, nothing else */
li r13, (TM_CAUSE_SYSCALL|TM_CAUSE_PERSISTENT)
TABORT(R13)
/*
* Return directly to userspace. We have corrupted user register state,
* but userspace will never see that register state. Execution will
* resume after the tbegin of the aborted transaction with the
* checkpointed register state.
*/
li r13, MSR_RI
andc r10, r10, r13
mtmsrd r10, 1
mtspr SPRN_SRR0, r11
mtspr SPRN_SRR1, r12
rfid
b . /* prevent speculative execution */
#endif
/* Save non-volatile GPRs, if not already saved. */
_GLOBAL(save_nvgprs)
ld r11,_TRAP(r1)
andi. r0,r11,1
beqlr-
SAVE_NVGPRS(r1)
clrrdi r0,r11,1
std r0,_TRAP(r1)
blr
[PATCH] syscall entry/exit revamp This cleanup patch speeds up the null syscall path on ppc64 by about 3%, and brings the ppc32 and ppc64 code slightly closer together. The ppc64 code was checking current_thread_info()->flags twice in the syscall exit path; once for TIF_SYSCALL_T_OR_A before disabling interrupts, and then again for TIF_SIGPENDING|TIF_NEED_RESCHED etc after disabling interrupts. Now we do the same as ppc32 -- check the flags only once in the fast path, and re-enable interrupts if necessary in the ptrace case. The patch abolishes the 'syscall_noerror' member of struct thread_info and replaces it with a TIF_NOERROR bit in the flags, which is handled in the slow path. This shortens the syscall entry code, which no longer needs to clear syscall_noerror. The patch adds a TIF_SAVE_NVGPRS flag which causes the syscall exit slow path to save the non-volatile GPRs into a signal frame. This removes the need for the assembly wrappers around sys_sigsuspend(), sys_rt_sigsuspend(), et al which existed solely to save those registers in advance. It also means I don't have to add new wrappers for ppoll() and pselect(), which is what I was supposed to be doing when I got distracted into this... Finally, it unifies the ppc64 and ppc32 methods of handling syscall exit directly into a signal handler (as required by sigsuspend et al) by introducing a TIF_RESTOREALL flag which causes _all_ the registers to be reloaded from the pt_regs by taking the ret_from_exception path, instead of the normal syscall exit path which stomps on the callee-saved GPRs. It appears to pass an LTP test run on ppc64, and passes basic testing on ppc32 too. Brief tests of ptrace functionality with strace and gdb also appear OK. I wouldn't send it to Linus for 2.6.15 just yet though :) Signed-off-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:52:18 +08:00
/*
* The sigsuspend and rt_sigsuspend system calls can call do_signal
* and thus put the process into the stopped state where we might
* want to examine its user state with ptrace. Therefore we need
* to save all the nonvolatile registers (r14 - r31) before calling
* the C code. Similarly, fork, vfork and clone need the full
* register state on the stack so that it can be copied to the child.
*/
_GLOBAL(ppc_fork)
bl save_nvgprs
bl sys_fork
b .Lsyscall_exit
_GLOBAL(ppc_vfork)
bl save_nvgprs
bl sys_vfork
b .Lsyscall_exit
_GLOBAL(ppc_clone)
bl save_nvgprs
bl sys_clone
b .Lsyscall_exit
powerpc: Fix various syscall/signal/swapcontext bugs A careful reading of the recent changes to the system call entry/exit paths revealed several problems, plus some things that could be simplified and improved: * 32-bit wasn't testing the _TIF_NOERROR bit in the syscall fast exit path, so it was only doing anything with it once it saw some other bit being set. In other words, the noerror behaviour would apply to the next system call where we had to reschedule or deliver a signal, which is not necessarily the current system call. * 32-bit wasn't doing the call to ptrace_notify in the syscall exit path when the _TIF_SINGLESTEP bit was set. * _TIF_RESTOREALL was in both _TIF_USER_WORK_MASK and _TIF_PERSYSCALL_MASK, which is odd since _TIF_RESTOREALL is only set by system calls. I took it out of _TIF_USER_WORK_MASK. * On 64-bit, _TIF_RESTOREALL wasn't causing the non-volatile registers to be restored (unless perhaps a signal was delivered or the syscall was traced or single-stepped). Thus the non-volatile registers weren't restored on exit from a signal handler. We probably got away with it mostly because signal handlers written in C wouldn't alter the non-volatile registers. * On 32-bit I simplified the code and made it more like 64-bit by making the syscall exit path jump to ret_from_except to handle preemption and signal delivery. * 32-bit was calling do_signal unnecessarily when _TIF_RESTOREALL was set - but I think because of that 32-bit was actually restoring the non-volatile registers on exit from a signal handler. * I changed the order of enabling interrupts and saving the non-volatile registers before calling do_syscall_trace_leave; now we enable interrupts first. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-03-08 10:24:22 +08:00
_GLOBAL(ppc32_swapcontext)
bl save_nvgprs
bl compat_sys_swapcontext
b .Lsyscall_exit
powerpc: Fix various syscall/signal/swapcontext bugs A careful reading of the recent changes to the system call entry/exit paths revealed several problems, plus some things that could be simplified and improved: * 32-bit wasn't testing the _TIF_NOERROR bit in the syscall fast exit path, so it was only doing anything with it once it saw some other bit being set. In other words, the noerror behaviour would apply to the next system call where we had to reschedule or deliver a signal, which is not necessarily the current system call. * 32-bit wasn't doing the call to ptrace_notify in the syscall exit path when the _TIF_SINGLESTEP bit was set. * _TIF_RESTOREALL was in both _TIF_USER_WORK_MASK and _TIF_PERSYSCALL_MASK, which is odd since _TIF_RESTOREALL is only set by system calls. I took it out of _TIF_USER_WORK_MASK. * On 64-bit, _TIF_RESTOREALL wasn't causing the non-volatile registers to be restored (unless perhaps a signal was delivered or the syscall was traced or single-stepped). Thus the non-volatile registers weren't restored on exit from a signal handler. We probably got away with it mostly because signal handlers written in C wouldn't alter the non-volatile registers. * On 32-bit I simplified the code and made it more like 64-bit by making the syscall exit path jump to ret_from_except to handle preemption and signal delivery. * 32-bit was calling do_signal unnecessarily when _TIF_RESTOREALL was set - but I think because of that 32-bit was actually restoring the non-volatile registers on exit from a signal handler. * I changed the order of enabling interrupts and saving the non-volatile registers before calling do_syscall_trace_leave; now we enable interrupts first. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-03-08 10:24:22 +08:00
_GLOBAL(ppc64_swapcontext)
bl save_nvgprs
bl sys_swapcontext
b .Lsyscall_exit
powerpc: Fix various syscall/signal/swapcontext bugs A careful reading of the recent changes to the system call entry/exit paths revealed several problems, plus some things that could be simplified and improved: * 32-bit wasn't testing the _TIF_NOERROR bit in the syscall fast exit path, so it was only doing anything with it once it saw some other bit being set. In other words, the noerror behaviour would apply to the next system call where we had to reschedule or deliver a signal, which is not necessarily the current system call. * 32-bit wasn't doing the call to ptrace_notify in the syscall exit path when the _TIF_SINGLESTEP bit was set. * _TIF_RESTOREALL was in both _TIF_USER_WORK_MASK and _TIF_PERSYSCALL_MASK, which is odd since _TIF_RESTOREALL is only set by system calls. I took it out of _TIF_USER_WORK_MASK. * On 64-bit, _TIF_RESTOREALL wasn't causing the non-volatile registers to be restored (unless perhaps a signal was delivered or the syscall was traced or single-stepped). Thus the non-volatile registers weren't restored on exit from a signal handler. We probably got away with it mostly because signal handlers written in C wouldn't alter the non-volatile registers. * On 32-bit I simplified the code and made it more like 64-bit by making the syscall exit path jump to ret_from_except to handle preemption and signal delivery. * 32-bit was calling do_signal unnecessarily when _TIF_RESTOREALL was set - but I think because of that 32-bit was actually restoring the non-volatile registers on exit from a signal handler. * I changed the order of enabling interrupts and saving the non-volatile registers before calling do_syscall_trace_leave; now we enable interrupts first. Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-03-08 10:24:22 +08:00
_GLOBAL(ppc_switch_endian)
bl save_nvgprs
bl sys_switch_endian
b .Lsyscall_exit
_GLOBAL(ret_from_fork)
bl schedule_tail
REST_NVGPRS(r1)
li r3,0
b .Lsyscall_exit
_GLOBAL(ret_from_kernel_thread)
bl schedule_tail
REST_NVGPRS(r1)
mtlr r14
mr r3,r15
#if defined(_CALL_ELF) && _CALL_ELF == 2
mr r12,r14
#endif
blrl
li r3,0
b .Lsyscall_exit
/*
* This routine switches between two different tasks. The process
* state of one is saved on its kernel stack. Then the state
* of the other is restored from its kernel stack. The memory
* management hardware is updated to the second process's state.
* Finally, we can return to the second process, via ret_from_except.
* On entry, r3 points to the THREAD for the current task, r4
* points to the THREAD for the new task.
*
* Note: there are two ways to get to the "going out" portion
* of this code; either by coming in via the entry (_switch)
* or via "fork" which must set up an environment equivalent
* to the "_switch" path. If you change this you'll have to change
* the fork code also.
*
* The code which creates the new task context is in 'copy_thread'
* in arch/powerpc/kernel/process.c
*/
.align 7
_GLOBAL(_switch)
mflr r0
std r0,16(r1)
stdu r1,-SWITCH_FRAME_SIZE(r1)
/* r3-r13 are caller saved -- Cort */
SAVE_8GPRS(14, r1)
SAVE_10GPRS(22, r1)
std r0,_NIP(r1) /* Return to switch caller */
mfcr r23
std r23,_CCR(r1)
std r1,KSP(r3) /* Set old stack pointer */
#ifdef CONFIG_SMP
/* We need a sync somewhere here to make sure that if the
* previous task gets rescheduled on another CPU, it sees all
* stores it has performed on this one.
*/
sync
#endif /* CONFIG_SMP */
powerpc: Feature nop out reservation clear when stcx checks address The POWER architecture does not require stcx to check that it is operating on the same address as the larx. This means it is possible for an an exception handler to execute a larx, get a reservation, decide not to do the stcx and then return back with an active reservation. If the interrupted code was in the middle of a larx/stcx sequence the stcx could incorrectly succeed. All recent POWER CPUs check the address before letting the stcx succeed so we can create a CPU feature and nop it out. As Ben suggested, we can only do this in our syscall path because there is a remote possibility some kernel code gets interrupted by an exception that ends up operating on the same cacheline. Thanks to Paul Mackerras and Derek Williams for the idea. To test this I used a very simple null syscall (actually getppid) testcase at http://ozlabs.org/~anton/junkcode/null_syscall.c I tested against 2.6.35-git10 with the following changes against the pseries_defconfig: CONFIG_VIRT_CPU_ACCOUNTING=n CONFIG_AUDIT=n CONFIG_PPC_4K_PAGES=n CONFIG_PPC_64K_PAGES=y CONFIG_FORCE_MAX_ZONEORDER=9 CONFIG_PPC_SUBPAGE_PROT=n CONFIG_FUNCTION_TRACER=n CONFIG_FUNCTION_GRAPH_TRACER=n CONFIG_IRQSOFF_TRACER=n CONFIG_STACK_TRACER=n to remove the overhead of virtual CPU accounting, syscall auditing and the ftrace mcount tracers. 64kB pages were enabled to minimise TLB misses. POWER6: +8.2% POWER7: +7.0% Another suggestion was to use a larx to something in the L1 instead of a stcx. This was almost as fast as removing the larx on POWER6, but only 3.5% faster on POWER7. We can use this to speed up the reservation clear in our exception exit code. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-08-11 09:40:27 +08:00
/*
* If we optimise away the clear of the reservation in system
* calls because we know the CPU tracks the address of the
* reservation, then we need to clear it here to cover the
* case that the kernel context switch path has no larx
* instructions.
*/
BEGIN_FTR_SECTION
ldarx r6,0,r1
END_FTR_SECTION_IFSET(CPU_FTR_STCX_CHECKS_ADDRESS)
#ifdef CONFIG_PPC_BOOK3S
/* Cancel all explict user streams as they will have no use after context
* switch and will stop the HW from creating streams itself
*/
DCBT_STOP_ALL_STREAM_IDS(r6)
#endif
addi r6,r4,-THREAD /* Convert THREAD to 'current' */
std r6,PACACURRENT(r13) /* Set new 'current' */
ld r8,KSP(r4) /* new stack pointer */
#ifdef CONFIG_PPC_BOOK3S
BEGIN_FTR_SECTION
clrrdi r6,r8,28 /* get its ESID */
clrrdi r9,r1,28 /* get current sp ESID */
FTR_SECTION_ELSE
clrrdi r6,r8,40 /* get its 1T ESID */
clrrdi r9,r1,40 /* get current sp 1T ESID */
ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_1T_SEGMENT)
clrldi. r0,r6,2 /* is new ESID c00000000? */
cmpd cr1,r6,r9 /* or is new ESID the same as current ESID? */
cror eq,4*cr1+eq,eq
beq 2f /* if yes, don't slbie it */
/* Bolt in the new stack SLB entry */
ld r7,KSP_VSID(r4) /* Get new stack's VSID */
oris r0,r6,(SLB_ESID_V)@h
ori r0,r0,(SLB_NUM_BOLTED-1)@l
BEGIN_FTR_SECTION
li r9,MMU_SEGSIZE_1T /* insert B field */
oris r6,r6,(MMU_SEGSIZE_1T << SLBIE_SSIZE_SHIFT)@h
rldimi r7,r9,SLB_VSID_SSIZE_SHIFT,0
END_MMU_FTR_SECTION_IFSET(MMU_FTR_1T_SEGMENT)
/* Update the last bolted SLB. No write barriers are needed
* here, provided we only update the current CPU's SLB shadow
* buffer.
*/
ld r9,PACA_SLBSHADOWPTR(r13)
li r12,0
std r12,SLBSHADOW_STACKESID(r9) /* Clear ESID */
li r12,SLBSHADOW_STACKVSID
STDX_BE r7,r12,r9 /* Save VSID */
li r12,SLBSHADOW_STACKESID
STDX_BE r0,r12,r9 /* Save ESID */
/* No need to check for MMU_FTR_NO_SLBIE_B here, since when
* we have 1TB segments, the only CPUs known to have the errata
* only support less than 1TB of system memory and we'll never
* actually hit this code path.
*/
slbie r6
slbie r6 /* Workaround POWER5 < DD2.1 issue */
slbmte r7,r0
isync
2:
#endif /* !CONFIG_PPC_BOOK3S */
CURRENT_THREAD_INFO(r7, r8) /* base of new stack */
/* Note: this uses SWITCH_FRAME_SIZE rather than INT_FRAME_SIZE
because we don't need to leave the 288-byte ABI gap at the
top of the kernel stack. */
addi r7,r7,THREAD_SIZE-SWITCH_FRAME_SIZE
mr r1,r8 /* start using new stack pointer */
std r7,PACAKSAVE(r13)
ld r6,_CCR(r1)
mtcrf 0xFF,r6
/* r3-r13 are destroyed -- Cort */
REST_8GPRS(14, r1)
REST_10GPRS(22, r1)
/* convert old thread to its task_struct for return value */
addi r3,r3,-THREAD
ld r7,_NIP(r1) /* Return to _switch caller in new task */
mtlr r7
addi r1,r1,SWITCH_FRAME_SIZE
blr
.align 7
_GLOBAL(ret_from_except)
ld r11,_TRAP(r1)
andi. r0,r11,1
bne ret_from_except_lite
REST_NVGPRS(r1)
_GLOBAL(ret_from_except_lite)
/*
* Disable interrupts so that current_thread_info()->flags
* can't change between when we test it and when we return
* from the interrupt.
*/
#ifdef CONFIG_PPC_BOOK3E
wrteei 0
#else
ld r10,PACAKMSR(r13) /* Get kernel MSR without EE */
mtmsrd r10,1 /* Update machine state */
#endif /* CONFIG_PPC_BOOK3E */
CURRENT_THREAD_INFO(r9, r1)
ld r3,_MSR(r1)
#ifdef CONFIG_PPC_BOOK3E
ld r10,PACACURRENT(r13)
#endif /* CONFIG_PPC_BOOK3E */
ld r4,TI_FLAGS(r9)
andi. r3,r3,MSR_PR
beq resume_kernel
#ifdef CONFIG_PPC_BOOK3E
lwz r3,(THREAD+THREAD_DBCR0)(r10)
#endif /* CONFIG_PPC_BOOK3E */
/* Check current_thread_info()->flags */
andi. r0,r4,_TIF_USER_WORK_MASK
bne 1f
powerpc: Restore FPU/VEC/VSX if previously used Currently the FPU, VEC and VSX facilities are lazily loaded. This is not a problem unless a process is using these facilities. Modern versions of GCC are very good at automatically vectorising code, new and modernised workloads make use of floating point and vector facilities, even the kernel makes use of vectorised memcpy. All this combined greatly increases the cost of a syscall since the kernel uses the facilities sometimes even in syscall fast-path making it increasingly common for a thread to take an *_unavailable exception soon after a syscall, not to mention potentially taking all three. The obvious overcompensation to this problem is to simply always load all the facilities on every exit to userspace. Loading up all FPU, VEC and VSX registers every time can be expensive and if a workload does avoid using them, it should not be forced to incur this penalty. An 8bit counter is used to detect if the registers have been used in the past and the registers are always loaded until the value wraps to back to zero. Several versions of the assembly in entry_64.S were tested: 1. Always calling C. 2. Performing a common case check and then calling C. 3. A complex check in asm. After some benchmarking it was determined that avoiding C in the common case is a performance benefit (option 2). The full check in asm (option 3) greatly complicated that codepath for a negligible performance gain and the trade-off was deemed not worth it. Signed-off-by: Cyril Bur <cyrilbur@gmail.com> [mpe: Move load_vec in the struct to fill an existing hole, reword change log] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> fixup
2016-02-29 14:53:47 +08:00
#ifdef CONFIG_PPC_BOOK3E
/*
* Check to see if the dbcr0 register is set up to debug.
* Use the internal debug mode bit to do this.
*/
andis. r0,r3,DBCR0_IDM@h
beq restore
mfmsr r0
rlwinm r0,r0,0,~MSR_DE /* Clear MSR.DE */
mtmsr r0
mtspr SPRN_DBCR0,r3
li r10, -1
mtspr SPRN_DBSR,r10
b restore
#else
powerpc: Restore FPU/VEC/VSX if previously used Currently the FPU, VEC and VSX facilities are lazily loaded. This is not a problem unless a process is using these facilities. Modern versions of GCC are very good at automatically vectorising code, new and modernised workloads make use of floating point and vector facilities, even the kernel makes use of vectorised memcpy. All this combined greatly increases the cost of a syscall since the kernel uses the facilities sometimes even in syscall fast-path making it increasingly common for a thread to take an *_unavailable exception soon after a syscall, not to mention potentially taking all three. The obvious overcompensation to this problem is to simply always load all the facilities on every exit to userspace. Loading up all FPU, VEC and VSX registers every time can be expensive and if a workload does avoid using them, it should not be forced to incur this penalty. An 8bit counter is used to detect if the registers have been used in the past and the registers are always loaded until the value wraps to back to zero. Several versions of the assembly in entry_64.S were tested: 1. Always calling C. 2. Performing a common case check and then calling C. 3. A complex check in asm. After some benchmarking it was determined that avoiding C in the common case is a performance benefit (option 2). The full check in asm (option 3) greatly complicated that codepath for a negligible performance gain and the trade-off was deemed not worth it. Signed-off-by: Cyril Bur <cyrilbur@gmail.com> [mpe: Move load_vec in the struct to fill an existing hole, reword change log] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> fixup
2016-02-29 14:53:47 +08:00
addi r3,r1,STACK_FRAME_OVERHEAD
bl restore_math
b restore
#endif
1: andi. r0,r4,_TIF_NEED_RESCHED
beq 2f
bl restore_interrupts
SCHEDULE_USER
b ret_from_except_lite
powerpc: Don't corrupt transactional state when using FP/VMX in kernel Currently, when we have a process using the transactional memory facilities on POWER8 (that is, the processor is in transactional or suspended state), and the process enters the kernel and the kernel then uses the floating-point or vector (VMX/Altivec) facility, we end up corrupting the user-visible FP/VMX/VSX state. This happens, for example, if a page fault causes a copy-on-write operation, because the copy_page function will use VMX to do the copy on POWER8. The test program below demonstrates the bug. The bug happens because when FP/VMX state for a transactional process is stored in the thread_struct, we store the checkpointed state in .fp_state/.vr_state and the transactional (current) state in .transact_fp/.transact_vr. However, when the kernel wants to use FP/VMX, it calls enable_kernel_fp() or enable_kernel_altivec(), which saves the current state in .fp_state/.vr_state. Furthermore, when we return to the user process we return with FP/VMX/VSX disabled. The next time the process uses FP/VMX/VSX, we don't know which set of state (the current register values, .fp_state/.vr_state, or .transact_fp/.transact_vr) we should be using, since we have no way to tell if we are still in the same transaction, and if not, whether the previous transaction succeeded or failed. Thus it is necessary to strictly adhere to the rule that if FP has been enabled at any point in a transaction, we must keep FP enabled for the user process with the current transactional state in the FP registers, until we detect that it is no longer in a transaction. Similarly for VMX; once enabled it must stay enabled until the process is no longer transactional. In order to keep this rule, we add a new thread_info flag which we test when returning from the kernel to userspace, called TIF_RESTORE_TM. This flag indicates that there is FP/VMX/VSX state to be restored before entering userspace, and when it is set the .tm_orig_msr field in the thread_struct indicates what state needs to be restored. The restoration is done by restore_tm_state(). The TIF_RESTORE_TM bit is set by new giveup_fpu/altivec_maybe_transactional helpers, which are called from enable_kernel_fp/altivec, giveup_vsx, and flush_fp/altivec_to_thread instead of giveup_fpu/altivec. The other thing to be done is to get the transactional FP/VMX/VSX state from .fp_state/.vr_state when doing reclaim, if that state has been saved there by giveup_fpu/altivec_maybe_transactional. Having done this, we set the FP/VMX bit in the thread's MSR after reclaim to indicate that that part of the state is now valid (having been reclaimed from the processor's checkpointed state). Finally, in the signal handling code, we move the clearing of the transactional state bits in the thread's MSR a bit earlier, before calling flush_fp_to_thread(), so that we don't unnecessarily set the TIF_RESTORE_TM bit. This is the test program: /* Michael Neuling 4/12/2013 * * See if the altivec state is leaked out of an aborted transaction due to * kernel vmx copy loops. * * gcc -m64 htm_vmxcopy.c -o htm_vmxcopy * */ /* We don't use all of these, but for reference: */ int main(int argc, char *argv[]) { long double vecin = 1.3; long double vecout; unsigned long pgsize = getpagesize(); int i; int fd; int size = pgsize*16; char tmpfile[] = "/tmp/page_faultXXXXXX"; char buf[pgsize]; char *a; uint64_t aborted = 0; fd = mkstemp(tmpfile); assert(fd >= 0); memset(buf, 0, pgsize); for (i = 0; i < size; i += pgsize) assert(write(fd, buf, pgsize) == pgsize); unlink(tmpfile); a = mmap(NULL, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0); assert(a != MAP_FAILED); asm __volatile__( "lxvd2x 40,0,%[vecinptr] ; " // set 40 to initial value TBEGIN "beq 3f ;" TSUSPEND "xxlxor 40,40,40 ; " // set 40 to 0 "std 5, 0(%[map]) ;" // cause kernel vmx copy page TABORT TRESUME TEND "li %[res], 0 ;" "b 5f ;" "3: ;" // Abort handler "li %[res], 1 ;" "5: ;" "stxvd2x 40,0,%[vecoutptr] ; " : [res]"=r"(aborted) : [vecinptr]"r"(&vecin), [vecoutptr]"r"(&vecout), [map]"r"(a) : "memory", "r0", "r3", "r4", "r5", "r6", "r7"); if (aborted && (vecin != vecout)){ printf("FAILED: vector state leaked on abort %f != %f\n", (double)vecin, (double)vecout); exit(1); } munmap(a, size); close(fd); printf("PASSED!\n"); return 0; } Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2014-01-13 12:56:29 +08:00
2:
#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
andi. r0,r4,_TIF_USER_WORK_MASK & ~_TIF_RESTORE_TM
bne 3f /* only restore TM if nothing else to do */
addi r3,r1,STACK_FRAME_OVERHEAD
bl restore_tm_state
powerpc: Don't corrupt transactional state when using FP/VMX in kernel Currently, when we have a process using the transactional memory facilities on POWER8 (that is, the processor is in transactional or suspended state), and the process enters the kernel and the kernel then uses the floating-point or vector (VMX/Altivec) facility, we end up corrupting the user-visible FP/VMX/VSX state. This happens, for example, if a page fault causes a copy-on-write operation, because the copy_page function will use VMX to do the copy on POWER8. The test program below demonstrates the bug. The bug happens because when FP/VMX state for a transactional process is stored in the thread_struct, we store the checkpointed state in .fp_state/.vr_state and the transactional (current) state in .transact_fp/.transact_vr. However, when the kernel wants to use FP/VMX, it calls enable_kernel_fp() or enable_kernel_altivec(), which saves the current state in .fp_state/.vr_state. Furthermore, when we return to the user process we return with FP/VMX/VSX disabled. The next time the process uses FP/VMX/VSX, we don't know which set of state (the current register values, .fp_state/.vr_state, or .transact_fp/.transact_vr) we should be using, since we have no way to tell if we are still in the same transaction, and if not, whether the previous transaction succeeded or failed. Thus it is necessary to strictly adhere to the rule that if FP has been enabled at any point in a transaction, we must keep FP enabled for the user process with the current transactional state in the FP registers, until we detect that it is no longer in a transaction. Similarly for VMX; once enabled it must stay enabled until the process is no longer transactional. In order to keep this rule, we add a new thread_info flag which we test when returning from the kernel to userspace, called TIF_RESTORE_TM. This flag indicates that there is FP/VMX/VSX state to be restored before entering userspace, and when it is set the .tm_orig_msr field in the thread_struct indicates what state needs to be restored. The restoration is done by restore_tm_state(). The TIF_RESTORE_TM bit is set by new giveup_fpu/altivec_maybe_transactional helpers, which are called from enable_kernel_fp/altivec, giveup_vsx, and flush_fp/altivec_to_thread instead of giveup_fpu/altivec. The other thing to be done is to get the transactional FP/VMX/VSX state from .fp_state/.vr_state when doing reclaim, if that state has been saved there by giveup_fpu/altivec_maybe_transactional. Having done this, we set the FP/VMX bit in the thread's MSR after reclaim to indicate that that part of the state is now valid (having been reclaimed from the processor's checkpointed state). Finally, in the signal handling code, we move the clearing of the transactional state bits in the thread's MSR a bit earlier, before calling flush_fp_to_thread(), so that we don't unnecessarily set the TIF_RESTORE_TM bit. This is the test program: /* Michael Neuling 4/12/2013 * * See if the altivec state is leaked out of an aborted transaction due to * kernel vmx copy loops. * * gcc -m64 htm_vmxcopy.c -o htm_vmxcopy * */ /* We don't use all of these, but for reference: */ int main(int argc, char *argv[]) { long double vecin = 1.3; long double vecout; unsigned long pgsize = getpagesize(); int i; int fd; int size = pgsize*16; char tmpfile[] = "/tmp/page_faultXXXXXX"; char buf[pgsize]; char *a; uint64_t aborted = 0; fd = mkstemp(tmpfile); assert(fd >= 0); memset(buf, 0, pgsize); for (i = 0; i < size; i += pgsize) assert(write(fd, buf, pgsize) == pgsize); unlink(tmpfile); a = mmap(NULL, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0); assert(a != MAP_FAILED); asm __volatile__( "lxvd2x 40,0,%[vecinptr] ; " // set 40 to initial value TBEGIN "beq 3f ;" TSUSPEND "xxlxor 40,40,40 ; " // set 40 to 0 "std 5, 0(%[map]) ;" // cause kernel vmx copy page TABORT TRESUME TEND "li %[res], 0 ;" "b 5f ;" "3: ;" // Abort handler "li %[res], 1 ;" "5: ;" "stxvd2x 40,0,%[vecoutptr] ; " : [res]"=r"(aborted) : [vecinptr]"r"(&vecin), [vecoutptr]"r"(&vecout), [map]"r"(a) : "memory", "r0", "r3", "r4", "r5", "r6", "r7"); if (aborted && (vecin != vecout)){ printf("FAILED: vector state leaked on abort %f != %f\n", (double)vecin, (double)vecout); exit(1); } munmap(a, size); close(fd); printf("PASSED!\n"); return 0; } Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2014-01-13 12:56:29 +08:00
b restore
3:
#endif
bl save_nvgprs
/*
* Use a non volatile GPR to save and restore our thread_info flags
* across the call to restore_interrupts.
*/
mr r30,r4
bl restore_interrupts
mr r4,r30
addi r3,r1,STACK_FRAME_OVERHEAD
bl do_notify_resume
b ret_from_except
resume_kernel:
/* check current_thread_info, _TIF_EMULATE_STACK_STORE */
andis. r8,r4,_TIF_EMULATE_STACK_STORE@h
beq+ 1f
addi r8,r1,INT_FRAME_SIZE /* Get the kprobed function entry */
lwz r3,GPR1(r1)
subi r3,r3,INT_FRAME_SIZE /* dst: Allocate a trampoline exception frame */
mr r4,r1 /* src: current exception frame */
mr r1,r3 /* Reroute the trampoline frame to r1 */
/* Copy from the original to the trampoline. */
li r5,INT_FRAME_SIZE/8 /* size: INT_FRAME_SIZE */
li r6,0 /* start offset: 0 */
mtctr r5
2: ldx r0,r6,r4
stdx r0,r6,r3
addi r6,r6,8
bdnz 2b
/* Do real store operation to complete stwu */
lwz r5,GPR1(r1)
std r8,0(r5)
/* Clear _TIF_EMULATE_STACK_STORE flag */
lis r11,_TIF_EMULATE_STACK_STORE@h
addi r5,r9,TI_FLAGS
0: ldarx r4,0,r5
andc r4,r4,r11
stdcx. r4,0,r5
bne- 0b
1:
#ifdef CONFIG_PREEMPT
/* Check if we need to preempt */
andi. r0,r4,_TIF_NEED_RESCHED
beq+ restore
/* Check that preempt_count() == 0 and interrupts are enabled */
lwz r8,TI_PREEMPT(r9)
cmpwi cr1,r8,0
ld r0,SOFTE(r1)
cmpdi r0,0
crandc eq,cr1*4+eq,eq
bne restore
/*
* Here we are preempting the current task. We want to make
* sure we are soft-disabled first and reconcile irq state.
*/
RECONCILE_IRQ_STATE(r3,r4)
1: bl preempt_schedule_irq
/* Re-test flags and eventually loop */
CURRENT_THREAD_INFO(r9, r1)
ld r4,TI_FLAGS(r9)
andi. r0,r4,_TIF_NEED_RESCHED
bne 1b
/*
* arch_local_irq_restore() from preempt_schedule_irq above may
* enable hard interrupt but we really should disable interrupts
* when we return from the interrupt, and so that we don't get
* interrupted after loading SRR0/1.
*/
#ifdef CONFIG_PPC_BOOK3E
wrteei 0
#else
ld r10,PACAKMSR(r13) /* Get kernel MSR without EE */
mtmsrd r10,1 /* Update machine state */
#endif /* CONFIG_PPC_BOOK3E */
#endif /* CONFIG_PREEMPT */
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
.globl fast_exc_return_irq
fast_exc_return_irq:
restore:
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
/*
* This is the main kernel exit path. First we check if we
* are about to re-enable interrupts
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
*/
ld r5,SOFTE(r1)
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
lbz r6,PACASOFTIRQEN(r13)
cmpwi cr0,r5,0
beq restore_irq_off
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
/* We are enabling, were we already enabled ? Yes, just return */
cmpwi cr0,r6,1
beq cr0,do_restore
/*
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
* We are about to soft-enable interrupts (we are hard disabled
* at this point). We check if there's anything that needs to
* be replayed first.
*/
lbz r0,PACAIRQHAPPENED(r13)
cmpwi cr0,r0,0
bne- restore_check_irq_replay
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
/*
* Get here when nothing happened while soft-disabled, just
* soft-enable and move-on. We will hard-enable as a side
* effect of rfi
*/
restore_no_replay:
TRACE_ENABLE_INTS
li r0,1
stb r0,PACASOFTIRQEN(r13);
/*
* Final return path. BookE is handled in a different file
*/
do_restore:
#ifdef CONFIG_PPC_BOOK3E
b exception_return_book3e
#else
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
/*
* Clear the reservation. If we know the CPU tracks the address of
* the reservation then we can potentially save some cycles and use
* a larx. On POWER6 and POWER7 this is significantly faster.
*/
BEGIN_FTR_SECTION
stdcx. r0,0,r1 /* to clear the reservation */
FTR_SECTION_ELSE
ldarx r4,0,r1
ALT_FTR_SECTION_END_IFCLR(CPU_FTR_STCX_CHECKS_ADDRESS)
/*
* Some code path such as load_up_fpu or altivec return directly
* here. They run entirely hard disabled and do not alter the
* interrupt state. They also don't use lwarx/stwcx. and thus
* are known not to leave dangling reservations.
*/
.globl fast_exception_return
fast_exception_return:
ld r3,_MSR(r1)
ld r4,_CTR(r1)
ld r0,_LINK(r1)
mtctr r4
mtlr r0
ld r4,_XER(r1)
mtspr SPRN_XER,r4
REST_8GPRS(5, r1)
andi. r0,r3,MSR_RI
beq- unrecov_restore
/* Load PPR from thread struct before we clear MSR:RI */
BEGIN_FTR_SECTION
ld r2,PACACURRENT(r13)
ld r2,TASKTHREADPPR(r2)
END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
/*
* Clear RI before restoring r13. If we are returning to
* userspace and we take an exception after restoring r13,
* we end up corrupting the userspace r13 value.
*/
ld r4,PACAKMSR(r13) /* Get kernel MSR without EE */
andc r4,r4,r0 /* r0 contains MSR_RI here */
mtmsrd r4,1
#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
/* TM debug */
std r3, PACATMSCRATCH(r13) /* Stash returned-to MSR */
#endif
/*
* r13 is our per cpu area, only restore it if we are returning to
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
* userspace the value stored in the stack frame may belong to
* another CPU.
*/
andi. r0,r3,MSR_PR
beq 1f
BEGIN_FTR_SECTION
mtspr SPRN_PPR,r2 /* Restore PPR */
END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
ACCOUNT_CPU_USER_EXIT(r2, r4)
REST_GPR(13, r1)
1:
mtspr SPRN_SRR1,r3
ld r2,_CCR(r1)
mtcrf 0xFF,r2
ld r2,_NIP(r1)
mtspr SPRN_SRR0,r2
ld r0,GPR0(r1)
ld r2,GPR2(r1)
ld r3,GPR3(r1)
ld r4,GPR4(r1)
ld r1,GPR1(r1)
rfid
b . /* prevent speculative execution */
#endif /* CONFIG_PPC_BOOK3E */
/*
* We are returning to a context with interrupts soft disabled.
*
* However, we may also about to hard enable, so we need to
* make sure that in this case, we also clear PACA_IRQ_HARD_DIS
* or that bit can get out of sync and bad things will happen
*/
restore_irq_off:
ld r3,_MSR(r1)
lbz r7,PACAIRQHAPPENED(r13)
andi. r0,r3,MSR_EE
beq 1f
rlwinm r7,r7,0,~PACA_IRQ_HARD_DIS
stb r7,PACAIRQHAPPENED(r13)
1: li r0,0
stb r0,PACASOFTIRQEN(r13);
TRACE_DISABLE_INTS
b do_restore
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
/*
* Something did happen, check if a re-emit is needed
* (this also clears paca->irq_happened)
*/
restore_check_irq_replay:
/* XXX: We could implement a fast path here where we check
* for irq_happened being just 0x01, in which case we can
* clear it and return. That means that we would potentially
* miss a decrementer having wrapped all the way around.
*
* Still, this might be useful for things like hash_page
*/
bl __check_irq_replay
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
cmpwi cr0,r3,0
beq restore_no_replay
/*
* We need to re-emit an interrupt. We do so by re-using our
* existing exception frame. We first change the trap value,
* but we need to ensure we preserve the low nibble of it
*/
ld r4,_TRAP(r1)
clrldi r4,r4,60
or r4,r4,r3
std r4,_TRAP(r1)
/*
* Then find the right handler and call it. Interrupts are
* still soft-disabled and we keep them that way.
*/
cmpwi cr0,r3,0x500
bne 1f
addi r3,r1,STACK_FRAME_OVERHEAD;
bl do_IRQ
b ret_from_except
1: cmpwi cr0,r3,0xe60
bne 1f
addi r3,r1,STACK_FRAME_OVERHEAD;
bl handle_hmi_exception
b ret_from_except
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
1: cmpwi cr0,r3,0x900
bne 1f
addi r3,r1,STACK_FRAME_OVERHEAD;
bl timer_interrupt
b ret_from_except
#ifdef CONFIG_PPC_DOORBELL
1:
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
#ifdef CONFIG_PPC_BOOK3E
cmpwi cr0,r3,0x280
#else
BEGIN_FTR_SECTION
cmpwi cr0,r3,0xe80
FTR_SECTION_ELSE
cmpwi cr0,r3,0xa00
ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE)
#endif /* CONFIG_PPC_BOOK3E */
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
bne 1f
addi r3,r1,STACK_FRAME_OVERHEAD;
bl doorbell_exception
b ret_from_except
#endif /* CONFIG_PPC_DOORBELL */
1: b ret_from_except /* What else to do here ? */
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
unrecov_restore:
addi r3,r1,STACK_FRAME_OVERHEAD
bl unrecoverable_exception
b unrecov_restore
#ifdef CONFIG_PPC_RTAS
/*
* On CHRP, the Run-Time Abstraction Services (RTAS) have to be
* called with the MMU off.
*
* In addition, we need to be in 32b mode, at least for now.
*
* Note: r3 is an input parameter to rtas, so don't trash it...
*/
_GLOBAL(enter_rtas)
mflr r0
std r0,16(r1)
stdu r1,-RTAS_FRAME_SIZE(r1) /* Save SP and create stack space. */
/* Because RTAS is running in 32b mode, it clobbers the high order half
* of all registers that it saves. We therefore save those registers
* RTAS might touch to the stack. (r0, r3-r13 are caller saved)
*/
SAVE_GPR(2, r1) /* Save the TOC */
SAVE_GPR(13, r1) /* Save paca */
SAVE_8GPRS(14, r1) /* Save the non-volatiles */
SAVE_10GPRS(22, r1) /* ditto */
mfcr r4
std r4,_CCR(r1)
mfctr r5
std r5,_CTR(r1)
mfspr r6,SPRN_XER
std r6,_XER(r1)
mfdar r7
std r7,_DAR(r1)
mfdsisr r8
std r8,_DSISR(r1)
/* Temporary workaround to clear CR until RTAS can be modified to
* ignore all bits.
*/
li r0,0
mtcr r0
#ifdef CONFIG_BUG
/* There is no way it is acceptable to get here with interrupts enabled,
* check it with the asm equivalent of WARN_ON
*/
lbz r0,PACASOFTIRQEN(r13)
1: tdnei r0,0
EMIT_BUG_ENTRY 1b,__FILE__,__LINE__,BUGFLAG_WARNING
#endif
/* Hard-disable interrupts */
mfmsr r6
rldicl r7,r6,48,1
rotldi r7,r7,16
mtmsrd r7,1
/* Unfortunately, the stack pointer and the MSR are also clobbered,
* so they are saved in the PACA which allows us to restore
* our original state after RTAS returns.
*/
std r1,PACAR1(r13)
std r6,PACASAVEDMSR(r13)
/* Setup our real return addr */
LOAD_REG_ADDR(r4,rtas_return_loc)
clrldi r4,r4,2 /* convert to realmode address */
mtlr r4
li r0,0
ori r0,r0,MSR_EE|MSR_SE|MSR_BE|MSR_RI
andc r0,r6,r0
li r9,1
rldicr r9,r9,MSR_SF_LG,(63-MSR_SF_LG)
ori r9,r9,MSR_IR|MSR_DR|MSR_FE0|MSR_FE1|MSR_FP|MSR_RI|MSR_LE
andc r6,r0,r9
sync /* disable interrupts so SRR0/1 */
mtmsrd r0 /* don't get trashed */
LOAD_REG_ADDR(r4, rtas)
ld r5,RTASENTRY(r4) /* get the rtas->entry value */
ld r4,RTASBASE(r4) /* get the rtas->base value */
mtspr SPRN_SRR0,r5
mtspr SPRN_SRR1,r6
rfid
b . /* prevent speculative execution */
rtas_return_loc:
FIXUP_ENDIAN
/* relocation is off at this point */
GET_PACA(r4)
clrldi r4,r4,2 /* convert to realmode address */
bcl 20,31,$+4
0: mflr r3
ld r3,(1f-0b)(r3) /* get &rtas_restore_regs */
mfmsr r6
li r0,MSR_RI
andc r6,r6,r0
sync
mtmsrd r6
ld r1,PACAR1(r4) /* Restore our SP */
ld r4,PACASAVEDMSR(r4) /* Restore our MSR */
mtspr SPRN_SRR0,r3
mtspr SPRN_SRR1,r4
rfid
b . /* prevent speculative execution */
.align 3
1: .llong rtas_restore_regs
rtas_restore_regs:
/* relocation is on at this point */
REST_GPR(2, r1) /* Restore the TOC */
REST_GPR(13, r1) /* Restore paca */
REST_8GPRS(14, r1) /* Restore the non-volatiles */
REST_10GPRS(22, r1) /* ditto */
GET_PACA(r13)
ld r4,_CCR(r1)
mtcr r4
ld r5,_CTR(r1)
mtctr r5
ld r6,_XER(r1)
mtspr SPRN_XER,r6
ld r7,_DAR(r1)
mtdar r7
ld r8,_DSISR(r1)
mtdsisr r8
addi r1,r1,RTAS_FRAME_SIZE /* Unstack our frame */
ld r0,16(r1) /* get return address */
mtlr r0
blr /* return to caller */
#endif /* CONFIG_PPC_RTAS */
_GLOBAL(enter_prom)
mflr r0
std r0,16(r1)
stdu r1,-PROM_FRAME_SIZE(r1) /* Save SP and create stack space */
/* Because PROM is running in 32b mode, it clobbers the high order half
* of all registers that it saves. We therefore save those registers
* PROM might touch to the stack. (r0, r3-r13 are caller saved)
*/
SAVE_GPR(2, r1)
SAVE_GPR(13, r1)
SAVE_8GPRS(14, r1)
SAVE_10GPRS(22, r1)
mfcr r10
mfmsr r11
std r10,_CCR(r1)
std r11,_MSR(r1)
/* Put PROM address in SRR0 */
mtsrr0 r4
/* Setup our trampoline return addr in LR */
bcl 20,31,$+4
0: mflr r4
addi r4,r4,(1f - 0b)
mtlr r4
/* Prepare a 32-bit mode big endian MSR
*/
#ifdef CONFIG_PPC_BOOK3E
rlwinm r11,r11,0,1,31
mtsrr1 r11
rfi
#else /* CONFIG_PPC_BOOK3E */
LOAD_REG_IMMEDIATE(r12, MSR_SF | MSR_ISF | MSR_LE)
andc r11,r11,r12
mtsrr1 r11
rfid
#endif /* CONFIG_PPC_BOOK3E */
1: /* Return from OF */
FIXUP_ENDIAN
/* Just make sure that r1 top 32 bits didn't get
* corrupt by OF
*/
rldicl r1,r1,0,32
/* Restore the MSR (back to 64 bits) */
ld r0,_MSR(r1)
MTMSRD(r0)
isync
/* Restore other registers */
REST_GPR(2, r1)
REST_GPR(13, r1)
REST_8GPRS(14, r1)
REST_10GPRS(22, r1)
ld r4,_CCR(r1)
mtcr r4
addi r1,r1,PROM_FRAME_SIZE
ld r0,16(r1)
mtlr r0
blr
#ifdef CONFIG_FUNCTION_TRACER
#ifdef CONFIG_DYNAMIC_FTRACE
_GLOBAL(mcount)
_GLOBAL(_mcount)
blr
_GLOBAL_TOC(ftrace_caller)
/* Taken from output of objdump from lib64/glibc */
mflr r3
ld r11, 0(r1)
stdu r1, -112(r1)
std r3, 128(r1)
ld r4, 16(r11)
subi r3, r3, MCOUNT_INSN_SIZE
.globl ftrace_call
ftrace_call:
bl ftrace_stub
nop
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
.globl ftrace_graph_call
ftrace_graph_call:
b ftrace_graph_stub
_GLOBAL(ftrace_graph_stub)
#endif
ld r0, 128(r1)
mtlr r0
addi r1, r1, 112
_GLOBAL(ftrace_stub)
blr
#else
_GLOBAL_TOC(_mcount)
/* Taken from output of objdump from lib64/glibc */
mflr r3
ld r11, 0(r1)
stdu r1, -112(r1)
std r3, 128(r1)
ld r4, 16(r11)
subi r3, r3, MCOUNT_INSN_SIZE
LOAD_REG_ADDR(r5,ftrace_trace_function)
ld r5,0(r5)
ld r5,0(r5)
mtctr r5
bctrl
nop
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
b ftrace_graph_caller
#endif
ld r0, 128(r1)
mtlr r0
addi r1, r1, 112
_GLOBAL(ftrace_stub)
blr
#endif /* CONFIG_DYNAMIC_FTRACE */
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
_GLOBAL(ftrace_graph_caller)
/* load r4 with local address */
ld r4, 128(r1)
subi r4, r4, MCOUNT_INSN_SIZE
/* Grab the LR out of the caller stack frame */
ld r11, 112(r1)
ld r3, 16(r11)
bl prepare_ftrace_return
nop
/*
* prepare_ftrace_return gives us the address we divert to.
* Change the LR in the callers stack frame to this.
*/
ld r11, 112(r1)
std r3, 16(r11)
ld r0, 128(r1)
mtlr r0
addi r1, r1, 112
blr
_GLOBAL(return_to_handler)
/* need to save return values */
std r4, -32(r1)
std r3, -24(r1)
/* save TOC */
std r2, -16(r1)
std r31, -8(r1)
mr r31, r1
stdu r1, -112(r1)
/*
* We might be called from a module.
* Switch to our TOC to run inside the core kernel.
*/
ld r2, PACATOC(r13)
bl ftrace_return_to_handler
nop
/* return value has real return address */
mtlr r3
ld r1, 0(r1)
ld r4, -32(r1)
ld r3, -24(r1)
ld r2, -16(r1)
ld r31, -8(r1)
/* Jump back to real return address */
blr
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
#endif /* CONFIG_FUNCTION_TRACER */