* [Qemu-devel] [PATCH] PXA-specific ARM hacks. @ 2007-03-16 22:06 andrzej zaborowski 2007-03-17 15:03 ` Paul Brook 0 siblings, 1 reply; 4+ messages in thread From: andrzej zaborowski @ 2007-03-16 22:06 UTC (permalink / raw) To: qemu-devel [-- Attachment #1: Type: text/plain, Size: 704 bytes --] This patch contains some ugly but very useful chnages: - in cpu-exec.c, halt the arm CPU on CPU_INTERRUPT_HALT like in other architectures. - wake the cpu up on CPU_INTERRUPT_EXITTB - this serves waking the CPU up without asserting IRQ or FIQ, which is possible on PXA. Would it be better to add a separate interrupt type instead? - in target-arm/translate.c, don't terminate the TB after a CP15 write. The purpose of this is to imitate the real processor's instruction cache (to some degree). OSes should never need rely on the cache prefetch but unfortunately Linux for pxa does when it's waking up from sleep or deep idle, when enabling the MMU (simplifies things a lot for Linux). Cheers, Andrew [-- Attachment #2: 0015-PXA-specific-ARM-hacks.txt --] [-- Type: text/plain, Size: 2379 bytes --] From 7ba76bd146394be87dc6480a172386a9e5e73877 Mon Sep 17 00:00:00 2001 From: Andrzej Zaborowski <balrog@zabor.org> Date: Fri, 16 Mar 2007 17:25:41 +0100 Subject: [PATCH] PXA-specific ARM hacks. --- cpu-all.h | 3 ++- cpu-exec.c | 10 ++++++++-- target-arm/translate.c | 2 ++ 3 files changed, 12 insertions(+), 3 deletions(-) diff --git a/cpu-all.h b/cpu-all.h index 9b617fc..177bcba 100644 --- a/cpu-all.h +++ b/cpu-all.h @@ -763,7 +763,8 @@ void cpu_dump_statistics (CPUState *env, FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...), int flags); -void cpu_abort(CPUState *env, const char *fmt, ...); +void cpu_abort(CPUState *env, const char *fmt, ...) + __attribute__ ((__format__ (__printf__, 2, 3))); extern CPUState *first_cpu; extern CPUState *cpu_single_env; extern int code_copy_enabled; diff --git a/cpu-exec.c b/cpu-exec.c index 48c2a93..2e99de2 100644 --- a/cpu-exec.c +++ b/cpu-exec.c @@ -276,8 +276,8 @@ int cpu_exec(CPUState *env1) if (env1->halted) { /* An interrupt wakes the CPU even if the I and F CPSR bits are set. */ - if (env1->interrupt_request - & (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD)) { + if (env1->interrupt_request & + (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD | CPU_INTERRUPT_EXITTB)) { env1->halted = 0; } else { return EXCP_HALTED; @@ -528,6 +528,12 @@ int cpu_exec(CPUState *env1) env->exception_index = EXCP_IRQ; do_interrupt(env); } + if (interrupt_request & CPU_INTERRUPT_HALT) { + env->interrupt_request &= ~CPU_INTERRUPT_HALT; + env->halted = 1; + env->exception_index = EXCP_HLT; + cpu_loop_exit(); + } #elif defined(TARGET_SH4) /* XXXXX */ #endif diff --git a/target-arm/translate.c b/target-arm/translate.c index e5032c5..c476e06 100644 --- a/target-arm/translate.c +++ b/target-arm/translate.c @@ -1536,7 +1536,9 @@ static int disas_cp15_insn(DisasContext *s, uint32_t insn) gen_movl_T0_reg(s, rd); gen_op_movl_cp15_T0(insn); } +#if 0 gen_lookup_tb(s); +#endif return 0; } -- 1.4.4.3 ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [Qemu-devel] [PATCH] PXA-specific ARM hacks. 2007-03-16 22:06 [Qemu-devel] [PATCH] PXA-specific ARM hacks andrzej zaborowski @ 2007-03-17 15:03 ` Paul Brook 2007-03-17 19:32 ` andrzej zaborowski 0 siblings, 1 reply; 4+ messages in thread From: Paul Brook @ 2007-03-17 15:03 UTC (permalink / raw) To: qemu-devel, balrogg On Friday 16 March 2007 22:06, andrzej zaborowski wrote: > This patch contains some ugly but very useful chnages: > - in cpu-exec.c, halt the arm CPU on CPU_INTERRUPT_HALT like in other > architectures. This is ok. Though I'd prefer it to actually be common code (like CPU_INTERRUPT_DEBUG). > - wake the cpu up on CPU_INTERRUPT_EXITTB - this serves waking the > CPU up without asserting IRQ or FIQ, which is possible on PXA. Would > it be better to add a separate interrupt type instead? I think this is probably ok. Could do with a comment somewhere saying so. > - in target-arm/translate.c, don't terminate the TB after a CP15 > write. The purpose of this is to imitate the real processor's > instruction cache (to some degree). OSes should never need rely on the > cache prefetch but unfortunately Linux for pxa does when it's waking > up from sleep or deep idle, when enabling the MMU (simplifies things a > lot for Linux). This is not acceptable in its current form. IIRC on arm cores the effects take effect once all insn have worked through the pipeline (typically a few cycles), and I bet there's code that relies on this. ie: mcr cp15, ...;nop;nop;nop;nop;@expect cp15 write to have taken effect by now A more acceptable solution would be: - Only extend the TB for the specific instruction that linux abuses - Limit the number of instructions that can follow before the end of the TB. - Add a comment saying what this hack is for. Paul ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Qemu-devel] [PATCH] PXA-specific ARM hacks. 2007-03-17 15:03 ` Paul Brook @ 2007-03-17 19:32 ` andrzej zaborowski 2007-03-17 20:16 ` andrzej zaborowski 0 siblings, 1 reply; 4+ messages in thread From: andrzej zaborowski @ 2007-03-17 19:32 UTC (permalink / raw) To: Paul Brook; +Cc: qemu-devel [-- Attachment #1: Type: text/plain, Size: 1743 bytes --] On 17/03/07, Paul Brook <paul@codesourcery.com> wrote: > On Friday 16 March 2007 22:06, andrzej zaborowski wrote: > > This patch contains some ugly but very useful chnages: > > - in cpu-exec.c, halt the arm CPU on CPU_INTERRUPT_HALT like in other > > architectures. > > This is ok. Though I'd prefer it to actually be common code (like > CPU_INTERRUPT_DEBUG). > > > - wake the cpu up on CPU_INTERRUPT_EXITTB - this serves waking the > > CPU up without asserting IRQ or FIQ, which is possible on PXA. Would > > it be better to add a separate interrupt type instead? > > I think this is probably ok. Could do with a comment somewhere saying so. > > > - in target-arm/translate.c, don't terminate the TB after a CP15 > > write. The purpose of this is to imitate the real processor's > > instruction cache (to some degree). OSes should never need rely on the > > cache prefetch but unfortunately Linux for pxa does when it's waking > > up from sleep or deep idle, when enabling the MMU (simplifies things a > > lot for Linux). > > This is not acceptable in its current form. IIRC on arm cores the effects take > effect once all insn have worked through the pipeline (typically a few > cycles), and I bet there's code that relies on this. ie: > mcr cp15, ...;nop;nop;nop;nop;@expect cp15 write to have taken effect by now > > A more acceptable solution would be: > - Only extend the TB for the specific instruction that linux abuses > - Limit the number of instructions that can follow before the end of the TB. > - Add a comment saying what this hack is for. Ok. Attaching a second version of this patch. The number of instructions after MMU enable and before TB ends is limited by the obligatory CPWAIT rather than in qemu. Regards, Andrzej [-- Attachment #2: 0015-PXA-specific-ARM-hacks.txt --] [-- Type: text/plain, Size: 3550 bytes --] From 37336004f611f544fc7d2eb763445fdca3109cdf Mon Sep 17 00:00:00 2001 From: Andrzej Zaborowski <balrog@zabor.org> Date: Sat, 17 Mar 2007 21:26:12 +0100 Subject: [PATCH] PXA-specific ARM hacks. --- cpu-all.h | 3 ++- cpu-exec.c | 20 +++++++++++--------- target-arm/translate.c | 7 ++++++- 3 files changed, 19 insertions(+), 11 deletions(-) diff --git a/cpu-all.h b/cpu-all.h index 9b617fc..177bcba 100644 --- a/cpu-all.h +++ b/cpu-all.h @@ -763,7 +763,8 @@ void cpu_dump_statistics (CPUState *env, FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...), int flags); -void cpu_abort(CPUState *env, const char *fmt, ...); +void cpu_abort(CPUState *env, const char *fmt, ...) + __attribute__ ((__format__ (__printf__, 2, 3))); extern CPUState *first_cpu; extern CPUState *cpu_single_env; extern int code_copy_enabled; diff --git a/cpu-exec.c b/cpu-exec.c index 48c2a93..10e2f2b 100644 --- a/cpu-exec.c +++ b/cpu-exec.c @@ -275,9 +275,10 @@ int cpu_exec(CPUState *env1) #elif defined(TARGET_ARM) if (env1->halted) { /* An interrupt wakes the CPU even if the I and F CPSR bits are - set. */ - if (env1->interrupt_request - & (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD)) { + set. We use EXITTB to silently wake CPU without causing an + actual interrupt. */ + if (env1->interrupt_request & + (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD | CPU_INTERRUPT_EXITTB)) { env1->halted = 0; } else { return EXCP_HALTED; @@ -414,6 +415,12 @@ int cpu_exec(CPUState *env1) env->exception_index = EXCP_DEBUG; cpu_loop_exit(); } + if (interrupt_request & CPU_INTERRUPT_HALT) { + env->interrupt_request &= ~CPU_INTERRUPT_HALT; + env->halted = 1; + env->exception_index = EXCP_HLT; + cpu_loop_exit(); + } #if defined(TARGET_I386) if ((interrupt_request & CPU_INTERRUPT_SMI) && !(env->hflags & HF_SMM_MASK)) { @@ -511,12 +518,7 @@ int cpu_exec(CPUState *env1) } else if (interrupt_request & CPU_INTERRUPT_TIMER) { //do_interrupt(0, 0, 0, 0, 0); env->interrupt_request &= ~CPU_INTERRUPT_TIMER; - } else if (interrupt_request & CPU_INTERRUPT_HALT) { - env->interrupt_request &= ~CPU_INTERRUPT_HALT; - env->halted = 1; - env->exception_index = EXCP_HLT; - cpu_loop_exit(); - } + } #elif defined(TARGET_ARM) if (interrupt_request & CPU_INTERRUPT_FIQ && !(env->uncached_cpsr & CPSR_F)) { diff --git a/target-arm/translate.c b/target-arm/translate.c index cd1c04d..885637c 100644 --- a/target-arm/translate.c +++ b/target-arm/translate.c @@ -1605,8 +1605,13 @@ static int disas_cp15_insn(DisasContext *s, uint32_t insn) } else { gen_movl_T0_reg(s, rd); gen_op_movl_cp15_T0(insn); + /* Normally we would always end the TB here, but Linux + * arch/arm/mach-pxa/sleep.S expects two instructions following + * an MMU enable to execute from cache. Imitate this behaviour. */ + if (!arm_feature(env, ARM_FEATURE_XSCALE) || + (insn & 0x0fff0fff) != 0x0e010f10) + gen_lookup_tb(s); } - gen_lookup_tb(s); return 0; } -- 1.4.4.3 ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [Qemu-devel] [PATCH] PXA-specific ARM hacks. 2007-03-17 19:32 ` andrzej zaborowski @ 2007-03-17 20:16 ` andrzej zaborowski 0 siblings, 0 replies; 4+ messages in thread From: andrzej zaborowski @ 2007-03-17 20:16 UTC (permalink / raw) To: Paul Brook; +Cc: qemu-devel [-- Attachment #1: Type: text/plain, Size: 1992 bytes --] On 17/03/07, andrzej zaborowski <balrog@zabor.org> wrote: > On 17/03/07, Paul Brook <paul@codesourcery.com> wrote: > > On Friday 16 March 2007 22:06, andrzej zaborowski wrote: > > > This patch contains some ugly but very useful chnages: > > > - in cpu-exec.c, halt the arm CPU on CPU_INTERRUPT_HALT like in other > > > architectures. > > > > This is ok. Though I'd prefer it to actually be common code (like > > CPU_INTERRUPT_DEBUG). > > > > > - wake the cpu up on CPU_INTERRUPT_EXITTB - this serves waking the > > > CPU up without asserting IRQ or FIQ, which is possible on PXA. Would > > > it be better to add a separate interrupt type instead? > > > > I think this is probably ok. Could do with a comment somewhere saying so. > > > > > - in target-arm/translate.c, don't terminate the TB after a CP15 > > > write. The purpose of this is to imitate the real processor's > > > instruction cache (to some degree). OSes should never need rely on the > > > cache prefetch but unfortunately Linux for pxa does when it's waking > > > up from sleep or deep idle, when enabling the MMU (simplifies things a > > > lot for Linux). > > > > This is not acceptable in its current form. IIRC on arm cores the effects take > > effect once all insn have worked through the pipeline (typically a few > > cycles), and I bet there's code that relies on this. ie: > > mcr cp15, ...;nop;nop;nop;nop;@expect cp15 write to have taken effect by now > > > > A more acceptable solution would be: > > - Only extend the TB for the specific instruction that linux abuses > > - Limit the number of instructions that can follow before the end of the TB. > > - Add a comment saying what this hack is for. > > Ok. Attaching a second version of this patch. The number of > instructions after MMU enable and before TB ends is limited by the > obligatory CPWAIT rather than in qemu. Oops, one more try. This time removing the use of cpu_reset() on entering Standby because cpu_reset() now does more than we want. Andrzej [-- Attachment #2: 0015-PXA-specific-ARM-hacks.txt --] [-- Type: text/plain, Size: 4781 bytes --] From 8d4d84ee71048638fd101ed27ba37b038ddf4651 Mon Sep 17 00:00:00 2001 From: Andrzej Zaborowski <balrog@zabor.org> Date: Sat, 17 Mar 2007 22:13:29 +0100 Subject: [PATCH] PXA-specific ARM hacks. --- cpu-all.h | 3 ++- cpu-exec.c | 20 +++++++++++--------- hw/pxa2xx.c | 3 ++- target-arm/translate.c | 11 ++++++++--- 4 files changed, 23 insertions(+), 14 deletions(-) diff --git a/cpu-all.h b/cpu-all.h index 9b617fc..177bcba 100644 --- a/cpu-all.h +++ b/cpu-all.h @@ -763,7 +763,8 @@ void cpu_dump_statistics (CPUState *env, FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...), int flags); -void cpu_abort(CPUState *env, const char *fmt, ...); +void cpu_abort(CPUState *env, const char *fmt, ...) + __attribute__ ((__format__ (__printf__, 2, 3))); extern CPUState *first_cpu; extern CPUState *cpu_single_env; extern int code_copy_enabled; diff --git a/cpu-exec.c b/cpu-exec.c index 48c2a93..10e2f2b 100644 --- a/cpu-exec.c +++ b/cpu-exec.c @@ -275,9 +275,10 @@ int cpu_exec(CPUState *env1) #elif defined(TARGET_ARM) if (env1->halted) { /* An interrupt wakes the CPU even if the I and F CPSR bits are - set. */ - if (env1->interrupt_request - & (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD)) { + set. We use EXITTB to silently wake CPU without causing an + actual interrupt. */ + if (env1->interrupt_request & + (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD | CPU_INTERRUPT_EXITTB)) { env1->halted = 0; } else { return EXCP_HALTED; @@ -414,6 +415,12 @@ int cpu_exec(CPUState *env1) env->exception_index = EXCP_DEBUG; cpu_loop_exit(); } + if (interrupt_request & CPU_INTERRUPT_HALT) { + env->interrupt_request &= ~CPU_INTERRUPT_HALT; + env->halted = 1; + env->exception_index = EXCP_HLT; + cpu_loop_exit(); + } #if defined(TARGET_I386) if ((interrupt_request & CPU_INTERRUPT_SMI) && !(env->hflags & HF_SMM_MASK)) { @@ -511,12 +518,7 @@ int cpu_exec(CPUState *env1) } else if (interrupt_request & CPU_INTERRUPT_TIMER) { //do_interrupt(0, 0, 0, 0, 0); env->interrupt_request &= ~CPU_INTERRUPT_TIMER; - } else if (interrupt_request & CPU_INTERRUPT_HALT) { - env->interrupt_request &= ~CPU_INTERRUPT_HALT; - env->halted = 1; - env->exception_index = EXCP_HLT; - cpu_loop_exit(); - } + } #elif defined(TARGET_ARM) if (interrupt_request & CPU_INTERRUPT_FIQ && !(env->uncached_cpsr & CPSR_F)) { diff --git a/hw/pxa2xx.c b/hw/pxa2xx.c index c8c6eb4..2d2d72a 100644 --- a/hw/pxa2xx.c +++ b/hw/pxa2xx.c @@ -259,7 +259,8 @@ static void pxa2xx_clkpwr_write(void *opaque, int op2, int reg, int crm, goto message; case 3: - cpu_reset(s->env); + s->env->uncached_cpsr = + ARM_CPU_MODE_SVC | CPSR_A | CPSR_F | CPSR_I; s->env->cp15.c1_sys = 0; s->env->cp15.c1_coproc = 0; s->env->cp15.c2 = 0; diff --git a/target-arm/translate.c b/target-arm/translate.c index cd1c04d..3e8367a 100644 --- a/target-arm/translate.c +++ b/target-arm/translate.c @@ -1579,7 +1579,7 @@ static int disas_cp_insn(DisasContext *s, uint32_t insn) /* Disassemble system coprocessor (cp15) instruction. Return nonzero if instruction is not defined. */ -static int disas_cp15_insn(DisasContext *s, uint32_t insn) +static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn) { uint32_t rd; @@ -1605,8 +1605,13 @@ static int disas_cp15_insn(DisasContext *s, uint32_t insn) } else { gen_movl_T0_reg(s, rd); gen_op_movl_cp15_T0(insn); + /* Normally we would always end the TB here, but Linux + * arch/arm/mach-pxa/sleep.S expects two instructions following + * an MMU enable to execute from cache. Imitate this behaviour. */ + if (!arm_feature(env, ARM_FEATURE_XSCALE) || + (insn & 0x0fff0fff) != 0x0e010f10) + gen_lookup_tb(s); } - gen_lookup_tb(s); return 0; } @@ -2907,7 +2912,7 @@ static void disas_arm_insn(CPUState * env, DisasContext *s) goto illegal_op; break; case 15: - if (disas_cp15_insn (s, insn)) + if (disas_cp15_insn (env, s, insn)) goto illegal_op; break; default: -- 1.4.4.3 ^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2007-03-17 20:17 UTC | newest] Thread overview: 4+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2007-03-16 22:06 [Qemu-devel] [PATCH] PXA-specific ARM hacks andrzej zaborowski 2007-03-17 15:03 ` Paul Brook 2007-03-17 19:32 ` andrzej zaborowski 2007-03-17 20:16 ` andrzej zaborowski
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).