RISC-V: Modularize common match conditions for trigger
cover RISC-V: Modularize common match conditions for trigger
According to RISC-V Debug specification, the enabled privilege levels of the trigger is common match conditions for all the types of the trigger. This series modularize the code for checking the privilege levels of type 2/3/6 triggers by implementing functions trigger_common_match() and trigger_priv_match().
Additional match conditions, such as CSR tcontrol and textra, can be further implemented into trigger_common_match() in the future.
Alvin Chang (4):
target/riscv: Add functions for common matching conditions of trigger target/riscv: Apply modularized matching conditions for breakpoint target/riscv: Apply modularized matching conditions for watchpoint target/riscv: Apply modularized matching conditions for icount trigger
target/riscv/debug.c | 124 +++++++++++++++++++++++++++++-------------- 1 file changed, 83 insertions(+), 41 deletions(-)
Signature · 1 line
2.34.1
PATCH 1/4 target/riscv: Add functions for common matching conditions of trigger
According to RISC-V Debug specification, there are several common matching conditions before firing a trigger, including the enabled privilege levels of the trigger.
This commit adds trigger_common_match() to prepare the common matching conditions for the type 2/3/6 triggers. For now, we just implement trigger_priv_match() to check if the enabled privilege levels of the trigger match CPU's current privilege level.
--- target/riscv/debug.c | 70 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 70 insertions(+)
target/riscv/debug.c +70 -0
@@ -241,6 +241,76 @@ static void do_trigger_action(CPURISCVState *env, target_ulong trigger_index) } } +/* +* Check the privilege level of specific trigger matches CPU's current privilege +* level. +*/ +static bool trigger_priv_match (CPURISCVState *env,trigger_type_t type, +int trigger_index) +{ +target_ulong ctrl = env->tdata1[trigger_index]; + +switch (type) { +case TRIGGER_TYPE_AD_MATCH : +/* type 2 trigger cannot be fired in VU/VS mode */ +if (env->virt_enabled) { +return false; + } +/* check U/S/M bit against current privilege level */ +if ((ctrl >>3 ) &BIT (env->priv)) { +return true; + } +break ; +case TRIGGER_TYPE_AD_MATCH6 : +if (env->virt_enabled) { +/* check VU/VS bit against current privilege level */ +if ((ctrl >>23 ) &BIT (env->priv)) { +return true; + } + }else { +/* check U/S/M bit against current privilege level */ +if ((ctrl >>3 ) &BIT (env->priv)) { +return true; + } + } +break ; +case TRIGGER_TYPE_INST_CNT : +if (env->virt_enabled) { +/* check VU/VS bit against current privilege level */ +if ((ctrl >>25 ) &BIT (env->priv)) { +return true; + } + }else { +/* check U/S/M bit against current privilege level */ +if ((ctrl >>6 ) &BIT (env->priv)) { +return true; + } + } +break ; +case TRIGGER_TYPE_INT : +case TRIGGER_TYPE_EXCP : +case TRIGGER_TYPE_EXT_SRC : +qemu_log_mask (LOG_UNIMP ,"trigger type: %d is not supported\n" , type); +break ; +case TRIGGER_TYPE_NO_EXIST : +case TRIGGER_TYPE_UNAVAIL : +qemu_log_mask (LOG_GUEST_ERROR ,"trigger type: %d does not exit\n" , + type); +break ; +default : +g_assert_not_reached (); + } + +return false; +} + +/* Common matching conditions for all types of the triggers. */ +static bool trigger_common_match (CPURISCVState *env,trigger_type_t type, +int trigger_index) +{ +return trigger_priv_match (env, type, trigger_index); +} +/* type 2 trigger */ static uint32_t type2_breakpoint_size (CPURISCVState *env,target_ulong ctrl)
Signature · 1 line
2.34.1
Re: [PATCH 1/4] target/riscv: Add functions for common matching conditions of trigger
Quoted 67 lines · Alvin Chang
According to RISC-V Debug specification, there are several common matching conditions before firing a trigger, including the enabled privilege levels of the trigger.
This commit adds
trigger_common_match()to prepare the common matching conditions for the type 2/3/6 triggers. For now, we just implementtrigger_priv_match()to check if the enabled privilege levels of the trigger match CPU's current privilege level.Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 70 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 70 insertions(+)target/riscv/debug.c +45
-0@@ -241,6 +241,76 @@ static void do_trigger_action(CPURISCVState *env, target_ulong trigger_index) } } +/* +* Check the privilege level of specific trigger matches CPU's current privilege +* level. +*/ +static bool trigger_priv_match (CPURISCVState *env,trigger_type_t type, +int trigger_index) +{ +target_ulong ctrl = env->tdata1[trigger_index]; + +switch (type) { +case TRIGGER_TYPE_AD_MATCH : +/* type 2 trigger cannot be fired in VU/VS mode */ +if (env->virt_enabled) { +return false; + } +/* check U/S/M bit against current privilege level */ +if ((ctrl >>3 ) &BIT (env->priv)) { +return true; + } +break ; +case TRIGGER_TYPE_AD_MATCH6 : +if (env->virt_enabled) { +/* check VU/VS bit against current privilege level */ +if ((ctrl >>23 ) &BIT (env->priv)) { +return true; + } + }else { +/* check U/S/M bit against current privilege level */ +if ((ctrl >>3 ) &BIT (env->priv)) { +return true; + } + } +break ; +case TRIGGER_TYPE_INST_CNT : +if (env->virt_enabled) { +/* check VU/VS bit against current privilege level */ +if ((ctrl >>25 ) &BIT (env->priv)) { +return true; + } + }else { +/* check U/S/M bit against current privilege level */ +if ((ctrl >>6 ) &BIT (env->priv)) { +return true; + } + }
target/riscv/debug.c static void do_trigger_action(CPURISCVState *env, target_ulong trigger_index) · line 298
Quoted 10 lines
target/riscv/debug.c +10
-0+break ; +case TRIGGER_TYPE_INT : +case TRIGGER_TYPE_EXCP : +case TRIGGER_TYPE_EXT_SRC : +qemu_log_mask (LOG_UNIMP ,"trigger type: %d is not supported\n" , type); +break ; +case TRIGGER_TYPE_NO_EXIST : +case TRIGGER_TYPE_UNAVAIL : +qemu_log_mask (LOG_GUEST_ERROR ,"trigger type: %d does not exit\n" , + type);
I believe you meant 'does not exist'.
Thanks,
Daniel
Quoted 18 lines
target/riscv/debug.c +15
-0+break ; +default : +g_assert_not_reached (); + } + +return false; +} + +/* Common matching conditions for all types of the triggers. */ +static bool trigger_common_match (CPURISCVState *env,trigger_type_t type, +int trigger_index) +{ +return trigger_priv_match (env, type, trigger_index); +} +/* type 2 trigger */ static uint32_t type2_breakpoint_size (CPURISCVState *env,target_ulong ctrl)
PATCH 2/4 target/riscv: Apply modularized matching conditions for breakpoint
In reply to Alvin Chang viaWe have implemented trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. Remove the related code in riscv_cpu_debug_check_breakpoint() and invoke trigger_common_match() to check the privilege levels of the type 2 and type 6 triggers for the breakpoints.
Only the execution bit and the executed PC should be futher checked in riscv_cpu_debug_check_breakpoint().
--- target/riscv/debug.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-)
target/riscv/debug.c +6 -20
@@ -855,21 +855,17 @@ bool riscv_cpu_debug_check_breakpoint(CPUState *cs)for (i =0 ; i <RV_MAX_TRIGGERS ; i++) { trigger_type =get_trigger_type (env, i); +if (!trigger_common_match (env, trigger_type, i)) { +continue ; + } +switch (trigger_type) {case TRIGGER_TYPE_AD_MATCH : -/* type 2 trigger cannot be fired in VU/VS mode */ -if (env->virt_enabled) { -return false; - } - ctrl = env->tdata1[i]; pc = env->tdata2[i];if ((ctrl &TYPE2_EXEC ) && (bp->pc == pc)) { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } +return true; }break ;case TRIGGER_TYPE_AD_MATCH6 :
@@ -877,17 +873,7 @@ bool riscv_cpu_debug_check_breakpoint(CPUState *cs) pc = env->tdata2[i];if ((ctrl &TYPE6_EXEC ) && (bp->pc == pc)) { -if (env->virt_enabled) { -/* check VU/VS bit against current privilege level */ -if ((ctrl >>23 ) &BIT (env->priv)) { -return true; - } - }else { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } - } +return true; }break ;default :
Signature · 1 line
2.34.1
Re: [PATCH 2/4] target/riscv: Apply modularized matching conditions for breakpoint
Quoted 7 lines · Alvin Chang
We have implemented
trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. Remove the related code inriscv_cpu_debug_check_breakpoint()and invoketrigger_common_match()to check the privilege levels of the type 2 and type 6 triggers for the breakpoints.Only the execution bit and the executed PC should be futher checked in
typo: further
Quoted 12 lines
riscv_cpu_debug_check_breakpoint().Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-)target/riscv/debug.c +0
-0
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs) · line 861
Quoted 7 lines
target/riscv/debug.c +4
-0@@ -855,21 +855,17 @@ bool riscv_cpu_debug_check_breakpoint(CPUState *cs)for (i =0 ; i <RV_MAX_TRIGGERS ; i++) { trigger_type =get_trigger_type (env, i); +if (!trigger_common_match (env, trigger_type, i)) { +continue ; + } +
I believe this will change how the function behaves. Before this patch, we would 'return false' if we have a TRIGGER_TYPE_AD_MATCH and env->virt_enabled is true.
Now, for the same case, we'll keep looping until we found a match, or return 'false' if we run out of triggers.
This seems ok to do, but we should state in the commit msg that we're intentionally change how the function works. If that's not the idea and we want to preserve the existing behavior, we would need to do:
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs) · line 860
Quoted 4 lines
target/riscv/debug.c +3
-0+if (!trigger_common_match (env, trigger_type, i)) { +return false; + }
Instead of just continue looping. Thanks,
Daniel
Quoted 38 lines
target/riscv/debug.c +2
-20switch (trigger_type) {case TRIGGER_TYPE_AD_MATCH : -/* type 2 trigger cannot be fired in VU/VS mode */ -if (env->virt_enabled) { -return false; - } - ctrl = env->tdata1[i]; pc = env->tdata2[i];if ((ctrl &TYPE2_EXEC ) && (bp->pc == pc)) { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } +return true; }break ;case TRIGGER_TYPE_AD_MATCH6 :@@ -877,17 +873,7 @@ bool riscv_cpu_debug_check_breakpoint(CPUState *cs) pc = env->tdata2[i];if ((ctrl &TYPE6_EXEC ) && (bp->pc == pc)) { -if (env->virt_enabled) { -/* check VU/VS bit against current privilege level */ -if ((ctrl >>23 ) &BIT (env->priv)) { -return true; - } - }else { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } - } +return true; }break ;default :
RE: [PATCH 2/4] target/riscv: Apply modularized matching conditions for breakpoint
Hi Daniel,
Quoted 20 lines
-----Original Message-----
From: Daniel Henrique Barboza <[email protected]>
Sent: Thursday, February 22, 2024 1:26 AM
To: Alvin Che-Chia Chang(張哲嘉) <[email protected]>; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH 2/4] target/riscv: Apply modularized matching conditions for breakpointQuoted 7 lines · Alvin Chang
We have implemented
trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. Remove the related code inriscv_cpu_debug_check_breakpoint()and invoketrigger_common_match()to check the privilege levels of the type 2 and type 6 triggers for the breakpoints.
Quoted 3 lines
Quoted 1 line
Only the execution bit and the executed PC should be futher checked in
typo: further
Thanks! Will fix it.
Quoted 22 lines
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState
Quoted 13 lines
riscv_cpu_debug_check_breakpoint().Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-)target/riscv/debug.c index renamed from target/riscv/debug.c +0
-07932233073..b971ed5d7a 100644
target/riscv/debug.c +0
-0@@ -855,21 +855,17 @@ bool riscv_cpu_debug_check_breakpoint(CPUState*cs)
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs) · line 861
Quoted 7 lines
target/riscv/debug.c +4
-0for (i =0 ; i <RV_MAX_TRIGGERS ; i++) { trigger_type =get_trigger_type (env, i); +if (!trigger_common_match (env, trigger_type, i)) { +continue ; + } +
Quoted 7 lines
I believe this will change how the function behaves. Before this patch, we would 'return false' if we have a
TRIGGER_TYPE_AD_MATCHandenv->virt_enabledis true.Now, for the same case, we'll keep looping until we found a match, or return 'false'
if we run out of triggers.
Oh! I didn't notice that the behavior is changed.. thank you.
Image that CPU supports both type 2 and type 6 triggers, and the debugger sets two identical breakpoints.(this should be a mistake, but we should not restrict the debugger) One of them is type 2 breakpoint at trigger index 0, and the other is type 6 breakpoint at trigger index 1.
Now if the system runs in VS/VU modes, it could match type 6 breakpoint, so the looping is necessary.
If the system runs in M/HS/U modes, it could match type 2 breakpoint.
Is my understanding correct?
Sincerely,
Alvin
Quoted 57 lines
This seems ok to do, but we should state in the commit msg that we're intentionally change how the function works. If that's not the idea and we want to preserve the existing behavior, we would need to do:
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs) · line 860
Quoted 4 lines
target/riscv/debug.c +3
-0+if (!trigger_common_match (env, trigger_type, i)) { +return false; + }Instead of just continue looping. Thanks,
Daniel
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs)
Quoted 12 lines
target/riscv/debug.c +0
-6switch (trigger_type) {case TRIGGER_TYPE_AD_MATCH : -/* type 2 trigger cannot be fired in VU/VS mode */ -if (env->virt_enabled) { -return false; - } - ctrl = env->tdata1[i]; pc = env->tdata2[i];if ((ctrl &TYPE2_EXEC ) && (bp->pc == pc)) { -/* check U/S/M bit against current privilege level */
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState
Quoted 8 lines
target/riscv/debug.c +1
-3-if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } +return true; }break ;case TRIGGER_TYPE_AD_MATCH6 :@@ -877,17 +873,7 @@ bool riscv_cpu_debug_check_breakpoint(CPUState*cs)
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs)
Quoted 5 lines
target/riscv/debug.c +0
-2pc = env->tdata2[i];if ((ctrl &TYPE6_EXEC ) && (bp->pc == pc)) { -if (env->virt_enabled) { -/* check VU/VS bit against current privilege level */
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs)
Quoted 5 lines
target/riscv/debug.c +0
-5-if ((ctrl >>23 ) &BIT (env->priv)) { -return true; - } - }else { -/* check U/S/M bit against current privilege level */
Quoted 8 lines
target/riscv/debug.c +1
-4-if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } - } +return true; }break ;default :
Re: [PATCH 2/4] target/riscv: Apply modularized matching conditions for breakpoint
Quoted 58 lines · Alvin Che-Chia Chang(張哲嘉)
Hi Daniel,
Quoted 20 lines
-----Original Message-----
From: Daniel Henrique Barboza <[email protected]>
Sent: Thursday, February 22, 2024 1:26 AM
To: Alvin Che-Chia Chang(張哲嘉) <[email protected]>; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH 2/4] target/riscv: Apply modularized matching conditions for breakpointQuoted 7 lines · Alvin Chang
We have implemented
trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. Remove the related code inriscv_cpu_debug_check_breakpoint()and invoketrigger_common_match()to check the privilege levels of the type 2 and type 6 triggers for the breakpoints.Quoted 3 lines
Quoted 1 line
Only the execution bit and the executed PC should be futher checked in
typo: further
Thanks! Will fix it.
Quoted 22 lines
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState
Quoted 13 lines
riscv_cpu_debug_check_breakpoint().Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-)target/riscv/debug.c index renamed from target/riscv/debug.c +0
-07932233073..b971ed5d7a 100644
target/riscv/debug.c +0
-0@@ -855,21 +855,17 @@ bool riscv_cpu_debug_check_breakpoint(CPUState*cs)
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs) · line 861
Quoted 7 lines
target/riscv/debug.c +4
-0for (i =0 ; i <RV_MAX_TRIGGERS ; i++) { trigger_type =get_trigger_type (env, i); +if (!trigger_common_match (env, trigger_type, i)) { +continue ; + } +Quoted 7 lines
I believe this will change how the function behaves. Before this patch, we would 'return false' if we have a
TRIGGER_TYPE_AD_MATCHandenv->virt_enabledis true.Now, for the same case, we'll keep looping until we found a match, or return 'false'
if we run out of triggers.
Quoted 7 lines
Oh! I didn't notice that the behavior is changed.. thank you.
Image that CPU supports both type 2 and type 6 triggers, and the debugger sets two identical breakpoints.(this should be a mistake, but we should not restrict the debugger) One of them is type 2 breakpoint at trigger index 0, and the other is type 6 breakpoint at trigger index 1.
Now if the system runs in VS/VU modes, it could match type 6 breakpoint, so the looping is necessary.
If the system runs in M/HS/U modes, it could match type 2 breakpoint.
Is my understanding correct?
It looks correct to me. We just need to mention in the commit msg that the behavior change is intentional, not an accident.
Thanks,
Daniel
Quoted 62 lines
Sincerely,
AlvinQuoted 57 lines
This seems ok to do, but we should state in the commit msg that we're intentionally change how the function works. If that's not the idea and we want to preserve the existing behavior, we would need to do:
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs) · line 860
Quoted 4 lines
target/riscv/debug.c +3
-0+if (!trigger_common_match (env, trigger_type, i)) { +return false; + }Instead of just continue looping. Thanks,
Daniel
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs)
Quoted 12 lines
target/riscv/debug.c +0
-6switch (trigger_type) {case TRIGGER_TYPE_AD_MATCH : -/* type 2 trigger cannot be fired in VU/VS mode */ -if (env->virt_enabled) { -return false; - } - ctrl = env->tdata1[i]; pc = env->tdata2[i];if ((ctrl &TYPE2_EXEC ) && (bp->pc == pc)) { -/* check U/S/M bit against current privilege level */
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState
Quoted 8 lines
target/riscv/debug.c +1
-3-if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } +return true; }break ;case TRIGGER_TYPE_AD_MATCH6 :@@ -877,17 +873,7 @@ bool riscv_cpu_debug_check_breakpoint(CPUState*cs)
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs)
Quoted 5 lines
target/riscv/debug.c +0
-2pc = env->tdata2[i];if ((ctrl &TYPE6_EXEC ) && (bp->pc == pc)) { -if (env->virt_enabled) { -/* check VU/VS bit against current privilege level */
target/riscv/debug.c bool riscv_cpu_debug_check_breakpoint(CPUState *cs)
Quoted 5 lines
target/riscv/debug.c +0
-5-if ((ctrl >>23 ) &BIT (env->priv)) { -return true; - } - }else { -/* check U/S/M bit against current privilege level */
Quoted 8 lines
target/riscv/debug.c +1
-4-if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } - } +return true; }break ;default :
PATCH 3/4 target/riscv: Apply modularized matching conditions for watchpoint
In reply to Alvin Chang viaWe have implemented trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. Remove the related code in riscv_cpu_debug_check_watchpoint() and invoke trigger_common_match() to check the privilege levels of the type 2 and type 6 triggers for the watchpoints.
Only load/store bits and loaded/stored address should be further checked in riscv_cpu_debug_check_watchpoint().
--- target/riscv/debug.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-)
target/riscv/debug.c +6 -20
@@ -899,13 +899,12 @@ bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp)for (i =0 ; i <RV_MAX_TRIGGERS ; i++) { trigger_type =get_trigger_type (env, i); +if (!trigger_common_match (env, trigger_type, i)) { +continue ; + } +switch (trigger_type) {case TRIGGER_TYPE_AD_MATCH : -/* type 2 trigger cannot be fired in VU/VS mode */ -if (env->virt_enabled) { -return false; - } - ctrl = env->tdata1[i]; addr = env->tdata2[i]; flags =0 ;
@@ -918,10 +917,7 @@ bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp) }if ((wp->flags & flags) && (wp->vaddr == addr)) { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } +return true; }break ;case TRIGGER_TYPE_AD_MATCH6 :
@@ -937,17 +933,7 @@ bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp) }if ((wp->flags & flags) && (wp->vaddr == addr)) { -if (env->virt_enabled) { -/* check VU/VS bit against current privilege level */ -if ((ctrl >>23 ) &BIT (env->priv)) { -return true; - } - }else { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } - } +return true; }break ;default :
Signature · 1 line
2.34.1
Re: [PATCH 3/4] target/riscv: Apply modularized matching conditions for watchpoint
Quoted 19 lines · Alvin Chang
We have implemented
trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. Remove the related code inriscv_cpu_debug_check_watchpoint()and invoketrigger_common_match()to check the privilege levels of the type 2 and type 6 triggers for the watchpoints.Only load/store bits and loaded/stored address should be further checked in
riscv_cpu_debug_check_watchpoint().Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-)target/riscv/debug.c +0
-0
target/riscv/debug.c bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp) · line 905
Quoted 7 lines
target/riscv/debug.c +4
-0@@ -899,13 +899,12 @@ bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp)for (i =0 ; i <RV_MAX_TRIGGERS ; i++) { trigger_type =get_trigger_type (env, i); +if (!trigger_common_match (env, trigger_type, i)) { +continue ; + } +
The same comments I made in patch 2 also applies here. It's ok to change how the function behaves as long as we're doing it on purpose and explaining why in the commit message. Otherwise this if (!trigger_common_match()" conditional should "return false" instead of keep looping to maintain the existing behavior.
Thanks,
Daniel
Quoted 41 lines
target/riscv/debug.c +2
-20switch (trigger_type) {case TRIGGER_TYPE_AD_MATCH : -/* type 2 trigger cannot be fired in VU/VS mode */ -if (env->virt_enabled) { -return false; - } - ctrl = env->tdata1[i]; addr = env->tdata2[i]; flags =0 ;@@ -918,10 +917,7 @@ bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp) }if ((wp->flags & flags) && (wp->vaddr == addr)) { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } +return true; }break ;case TRIGGER_TYPE_AD_MATCH6 :@@ -937,17 +933,7 @@ bool riscv_cpu_debug_check_watchpoint(CPUState *cs, CPUWatchpoint *wp) }if ((wp->flags & flags) && (wp->vaddr == addr)) { -if (env->virt_enabled) { -/* check VU/VS bit against current privilege level */ -if ((ctrl >>23 ) &BIT (env->priv)) { -return true; - } - }else { -/* check U/S/M bit against current privilege level */ -if ((ctrl >>3 ) &BIT (env->priv)) { -return true; - } - } +return true; }break ;default :
PATCH 4/4 target/riscv: Apply modularized matching conditions for icount trigger
In reply to Alvin Chang viaWe have implemented trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. We can invoke trigger_common_match() to check the privilege levels of the type 3 triggers.
--- target/riscv/debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/debug.c +1 -1
@@ -624,7 +624,7 @@ void helper_itrigger_match(CPURISCVState *env)if (get_trigger_type (env, i) !=TRIGGER_TYPE_INST_CNT ) {continue ; } -if (check_itrigger_priv (env, i)) { +if (!trigger_common_match (env,TRIGGER_TYPE_INST_CNT , i)) {continue ; } count =itrigger_get_count (env, i);
Signature · 1 line
2.34.1
Re: [PATCH 4/4] target/riscv: Apply modularized matching conditions for icount trigger
Quoted 15 lines · Alvin Chang
We have implemented
trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. We can invoketrigger_common_match()to check the privilege levels of the type 3 triggers.Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)target/riscv/debug.c +0
-0
target/riscv/debug.c void helper_itrigger_match(CPURISCVState *env) · line 629
Quoted 7 lines
target/riscv/debug.c +1
-1@@ -624,7 +624,7 @@ void helper_itrigger_match(CPURISCVState *env)if (get_trigger_type (env, i) !=TRIGGER_TYPE_INST_CNT ) {continue ; } -if (check_itrigger_priv (env, i)) { +if (!trigger_common_match (env,TRIGGER_TYPE_INST_CNT , i)) {continue ; }
Looks good. Shouldn't we also change riscv_itrigger_enabled() to also use trigger_common_match()? riscv_itrigger_enabled() is remarkably similar to helper_itrigger_match() so I believe we can also use the new function there.
Thanks,
Daniel
Quoted 1 line
target/riscv/debug.c +0
-0count =itrigger_get_count (env, i);
RE: [PATCH 4/4] target/riscv: Apply modularized matching conditions for icount trigger
Hi Daniel,
Quoted 32 lines
-----Original Message-----
From: Daniel Henrique Barboza <[email protected]>
Sent: Thursday, February 22, 2024 2:06 AM
To: Alvin Che-Chia Chang(張哲嘉) <[email protected]>; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH 4/4] target/riscv: Apply modularized matching conditions for icount triggerQuoted 19 lines · Alvin Chang
We have implemented
trigger_common_match(), which checks if the enabled privilege levels of the trigger match CPU's current privilege level. We can invoketrigger_common_match()to check the privilege levels of the type 3 triggers.Signed-off-by Alvin Chang <[email protected]>--- target/riscv/debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)target/riscv/debug.c index renamed from target/riscv/debug.c +0
-0
67ba19c966..de996a393c100644target/riscv/debug.c +0
-1@@ -624,7 +624,7 @@ void helper_itrigger_match(CPURISCVState *env)if (get_trigger_type (env, i) !=TRIGGER_TYPE_INST_CNT ) {continue ; } -if (check_itrigger_priv (env, i)) {
Quoted 8 lines
target/riscv/debug.c void helper_itrigger_match(CPURISCVState *env) · line 629
Quoted 3 lines
target/riscv/debug.c +1
-0+if (!trigger_common_match (env,TRIGGER_TYPE_INST_CNT , i)) {continue ; }Looks good. Shouldn't we also change
riscv_itrigger_enabled()to also usetrigger_common_match()?riscv_itrigger_enabled()is remarkably similar tohelper_itrigger_match()so I believe we can also use the new function there.
I think we might not want to apply trigger_common_match() into riscv_itrigger_enabled().
The trigger_common_match() is used to check if the trigger can be matched in current privilege level.
It will check many conditions: trigger privilege levels, textra, tcontrol, etc.
The riscv_itrigger_enabled() is used to check if any icount trigger is enabled by checking vs/vu/count/s/u fields of tdata1 only.
In fact, we found the comparisons between tdata1 bit-fields and env->priv in check_itrigger_priv() are bugs. And we have a patch to fix that:
bool riscv_itrigger_enabled(CPURISCVState *env)
{
int count;
for (int i = 0; i < RV_MAX_TRIGGERS; i++) {
if (get_trigger_type(env, i) != TRIGGER_TYPE_INST_CNT) {
continue;
}
- if (check_itrigger_priv(env, i)) {
+ if ((env->tdata1[i] & ITRIGGER_VS) == 0 &&
+ (env->tdata1[i] & ITRIGGER_VU) == 0 &&
+ (env->tdata1[i] & ITRIGGER_U) == 0 &&
+ (env->tdata1[i] & ITRIGGER_S) == 0 &&
+ (env->tdata1[i] & ITRIGGER_M) == 0 ) {
continue;
}
count = itrigger_get_count(env, i);
if (!count) {
continue;
}
return true;
}
return false;
}Sincerely,
Alvin Chang
Quoted 11 lines
Thanks,
Daniel
Quoted 1 line
target/riscv/debug.c +0
-0count =itrigger_get_count (env, i);