* [Qemu-devel] [PATCH for-2.12 0/4] Fix various BRK/BKPT related bugs
@ 2018-03-20 13:41 Peter Maydell
2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK Peter Maydell
` (3 more replies)
0 siblings, 4 replies; 11+ messages in thread
From: Peter Maydell @ 2018-03-20 13:41 UTC (permalink / raw)
To: qemu-arm, qemu-devel; +Cc: patches
The main aim of this patchset is to fix the bug reported in
https://bugs.launchpad.net/qemu/+bug/1756927 where we don't
report the correct FSR when a BKPT instruction causes us to
take an exception to AArch32 with LPAE enabled. The other
patches fix some minor bugs found along the way:
* we weren't honouring MDCR_EL2.TDE when delivering an
exception caused by a BRK or BKPT insn, so we'd always
send it to EL1 even if it should have gone to EL2
* our logic to decide whether to report a short-format or
long-format LPAE for hardware breakpoint and watchpoint
exceptions would get it wrong in some corner cases: we could
select the short format because we're at EL0 and the EL1
translation regime is not using LPAE, but then route the
debug exception to EL2 because of MDCR_EL2.TDE and hand
EL2 the wrong format FSR.
* when taking a debug exception in AArch32, we would set the
guest visible IFAR to whatever old value was lying around
in env->exception.vaddress. IFAR in this situation is
architecturally UNKNOWN, but letting it be possibly set
to an old value that the guest shouldn't be able to see
at its current security/exception level is not permitted,
and setting it to a non-deterministic value is bad for
record-and-replay.
(This patchset further reduces the set of places that are
calling the somewhat dubious extended_addresses_enabled()
function. The last callsite is in the CONTEXTIDR write
function, and that should probably be updated something along
the lines of the ideas I wrote in this email last year:
https://www.mail-archive.com/qemu-devel@nongnu.org/msg491671.html
but that is a cleanup for another day.)
thanks
-- PMM
Peter Maydell (4):
target/arm: Honour MDCR_EL2.TDE when routing exceptions due to
BKPT/BRK
target/arm: Factor out code to calculate FSR for debug exceptions
target/arm: Set FSR for BKPT, BRK when raising exception
target/arm: Always set FAR to a known unknown value for debug
exceptions
target/arm/helper.h | 1 +
target/arm/internals.h | 25 +++++++++++++++++++++++++
target/arm/helper.c | 1 -
target/arm/op_helper.c | 33 ++++++++++++++++++++++-----------
target/arm/translate-a64.c | 15 +++++++++++++--
target/arm/translate.c | 19 ++++++++++++++-----
6 files changed, 75 insertions(+), 19 deletions(-)
--
2.16.2
^ permalink raw reply [flat|nested] 11+ messages in thread* [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK 2018-03-20 13:41 [Qemu-devel] [PATCH for-2.12 0/4] Fix various BRK/BKPT related bugs Peter Maydell @ 2018-03-20 13:41 ` Peter Maydell 2018-03-22 8:22 ` Philippe Mathieu-Daudé 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions Peter Maydell ` (2 subsequent siblings) 3 siblings, 1 reply; 11+ messages in thread From: Peter Maydell @ 2018-03-20 13:41 UTC (permalink / raw) To: qemu-arm, qemu-devel; +Cc: patches The MDCR_EL2.TDE bit allows the exception level targeted by debug exceptions to be set to EL2 for code executing at EL0. We handle this in the arm_debug_target_el() function, but this is only used for hardware breakpoint and watchpoint exceptions, not for the exception generated when the guest executes an AArch32 BKPT or AArch64 BRK instruction. We don't have enough information for a translate-time equivalent of arm_debug_target_el(), so instead make BKPT and BRK call a special purpose helper which can do the routing, rather than the generic exception_with_syndrome helper. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 1 + target/arm/op_helper.c | 8 ++++++++ target/arm/translate-a64.c | 15 +++++++++++++-- target/arm/translate.c | 19 ++++++++++++++----- 4 files changed, 36 insertions(+), 7 deletions(-) diff --git a/target/arm/helper.h b/target/arm/helper.h index 0d2094f2be..34e8cc8904 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -47,6 +47,7 @@ DEF_HELPER_FLAGS_3(sel_flags, TCG_CALL_NO_RWG_SE, i32, i32, i32, i32) DEF_HELPER_2(exception_internal, void, env, i32) DEF_HELPER_4(exception_with_syndrome, void, env, i32, i32, i32) +DEF_HELPER_2(exception_bkpt_insn, void, env, i32) DEF_HELPER_1(setend, void, env) DEF_HELPER_2(wfi, void, env, i32) DEF_HELPER_1(wfe, void, env) diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c index 7a88fd2c92..4b123d2bd6 100644 --- a/target/arm/op_helper.c +++ b/target/arm/op_helper.c @@ -483,6 +483,14 @@ void HELPER(exception_with_syndrome)(CPUARMState *env, uint32_t excp, raise_exception(env, excp, syndrome, target_el); } +/* Raise an EXCP_BKPT with the specified syndrome register value, + * targeting the correct exception level for debug exceptions. + */ +void HELPER(exception_bkpt_insn)(CPUARMState *env, uint32_t syndrome) +{ + raise_exception(env, EXCP_BKPT, syndrome, arm_debug_target_el(env)); +} + uint32_t HELPER(cpsr_read)(CPUARMState *env) { return cpsr_read(env) & ~(CPSR_EXEC | CPSR_RESERVED); diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index 31ff0479e6..510951f7c7 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -321,6 +321,18 @@ static void gen_exception_insn(DisasContext *s, int offset, int excp, s->base.is_jmp = DISAS_NORETURN; } +static void gen_exception_bkpt_insn(DisasContext *s, int offset, + uint32_t syndrome) +{ + TCGv_i32 tcg_syn; + + gen_a64_set_pc_im(s->pc - offset); + tcg_syn = tcg_const_i32(syndrome); + gen_helper_exception_bkpt_insn(cpu_env, tcg_syn); + tcg_temp_free_i32(tcg_syn); + s->base.is_jmp = DISAS_NORETURN; +} + static void gen_ss_advance(DisasContext *s) { /* If the singlestep state is Active-not-pending, advance to @@ -1839,8 +1851,7 @@ static void disas_exc(DisasContext *s, uint32_t insn) break; } /* BRK */ - gen_exception_insn(s, 4, EXCP_BKPT, syn_aa64_bkpt(imm16), - default_exception_el(s)); + gen_exception_bkpt_insn(s, 4, syn_aa64_bkpt(imm16)); break; case 2: if (op2_ll != 0) { diff --git a/target/arm/translate.c b/target/arm/translate.c index ba6ab7d287..68f0c585f4 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -1248,6 +1248,18 @@ static void gen_exception_insn(DisasContext *s, int offset, int excp, s->base.is_jmp = DISAS_NORETURN; } +static void gen_exception_bkpt_insn(DisasContext *s, int offset, int syn) +{ + TCGv_i32 tcg_syn; + + gen_set_condexec(s); + gen_set_pc_im(s, s->pc - offset); + tcg_syn = tcg_const_i32(syn); + gen_helper_exception_bkpt_insn(cpu_env, tcg_syn); + tcg_temp_free_i32(tcg_syn); + s->base.is_jmp = DISAS_NORETURN; +} + /* Force a TB lookup after an instruction that changes the CPU state. */ static inline void gen_lookup_tb(DisasContext *s) { @@ -8774,9 +8786,7 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) case 1: /* bkpt */ ARCH(5); - gen_exception_insn(s, 4, EXCP_BKPT, - syn_aa32_bkpt(imm16, false), - default_exception_el(s)); + gen_exception_bkpt_insn(s, 4, syn_aa32_bkpt(imm16, false)); break; case 2: /* Hypervisor call (v7) */ @@ -11983,8 +11993,7 @@ static void disas_thumb_insn(DisasContext *s, uint32_t insn) { int imm8 = extract32(insn, 0, 8); ARCH(5); - gen_exception_insn(s, 2, EXCP_BKPT, syn_aa32_bkpt(imm8, true), - default_exception_el(s)); + gen_exception_bkpt_insn(s, 2, syn_aa32_bkpt(imm8, true)); break; } -- 2.16.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK Peter Maydell @ 2018-03-22 8:22 ` Philippe Mathieu-Daudé 0 siblings, 0 replies; 11+ messages in thread From: Philippe Mathieu-Daudé @ 2018-03-22 8:22 UTC (permalink / raw) To: Peter Maydell, qemu-arm, qemu-devel; +Cc: patches Hi Peter, On 03/20/2018 02:41 PM, Peter Maydell wrote: > The MDCR_EL2.TDE bit allows the exception level targeted by debug > exceptions to be set to EL2 for code executing at EL0. We handle > this in the arm_debug_target_el() function, but this is only used for > hardware breakpoint and watchpoint exceptions, not for the exception > generated when the guest executes an AArch32 BKPT or AArch64 BRK > instruction. We don't have enough information for a translate-time > equivalent of arm_debug_target_el(), so instead make BKPT and BRK > call a special purpose helper which can do the routing, rather than > the generic exception_with_syndrome helper. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > target/arm/helper.h | 1 + > target/arm/op_helper.c | 8 ++++++++ > target/arm/translate-a64.c | 15 +++++++++++++-- > target/arm/translate.c | 19 ++++++++++++++----- > 4 files changed, 36 insertions(+), 7 deletions(-) > > diff --git a/target/arm/helper.h b/target/arm/helper.h > index 0d2094f2be..34e8cc8904 100644 > --- a/target/arm/helper.h > +++ b/target/arm/helper.h > @@ -47,6 +47,7 @@ DEF_HELPER_FLAGS_3(sel_flags, TCG_CALL_NO_RWG_SE, > i32, i32, i32, i32) > DEF_HELPER_2(exception_internal, void, env, i32) > DEF_HELPER_4(exception_with_syndrome, void, env, i32, i32, i32) > +DEF_HELPER_2(exception_bkpt_insn, void, env, i32) > DEF_HELPER_1(setend, void, env) > DEF_HELPER_2(wfi, void, env, i32) > DEF_HELPER_1(wfe, void, env) > diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c > index 7a88fd2c92..4b123d2bd6 100644 > --- a/target/arm/op_helper.c > +++ b/target/arm/op_helper.c > @@ -483,6 +483,14 @@ void HELPER(exception_with_syndrome)(CPUARMState *env, uint32_t excp, > raise_exception(env, excp, syndrome, target_el); > } > > +/* Raise an EXCP_BKPT with the specified syndrome register value, > + * targeting the correct exception level for debug exceptions. > + */ > +void HELPER(exception_bkpt_insn)(CPUARMState *env, uint32_t syndrome) > +{ > + raise_exception(env, EXCP_BKPT, syndrome, arm_debug_target_el(env)); > +} > + > uint32_t HELPER(cpsr_read)(CPUARMState *env) > { > return cpsr_read(env) & ~(CPSR_EXEC | CPSR_RESERVED); > diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c > index 31ff0479e6..510951f7c7 100644 > --- a/target/arm/translate-a64.c > +++ b/target/arm/translate-a64.c > @@ -321,6 +321,18 @@ static void gen_exception_insn(DisasContext *s, int offset, int excp, > s->base.is_jmp = DISAS_NORETURN; > } > > +static void gen_exception_bkpt_insn(DisasContext *s, int offset, > + uint32_t syndrome) > +{ > + TCGv_i32 tcg_syn; > + > + gen_a64_set_pc_im(s->pc - offset); > + tcg_syn = tcg_const_i32(syndrome); > + gen_helper_exception_bkpt_insn(cpu_env, tcg_syn); > + tcg_temp_free_i32(tcg_syn); > + s->base.is_jmp = DISAS_NORETURN; > +} > + > static void gen_ss_advance(DisasContext *s) > { > /* If the singlestep state is Active-not-pending, advance to > @@ -1839,8 +1851,7 @@ static void disas_exc(DisasContext *s, uint32_t insn) > break; > } > /* BRK */ > - gen_exception_insn(s, 4, EXCP_BKPT, syn_aa64_bkpt(imm16), > - default_exception_el(s)); > + gen_exception_bkpt_insn(s, 4, syn_aa64_bkpt(imm16)); > break; > case 2: > if (op2_ll != 0) { > diff --git a/target/arm/translate.c b/target/arm/translate.c > index ba6ab7d287..68f0c585f4 100644 > --- a/target/arm/translate.c > +++ b/target/arm/translate.c > @@ -1248,6 +1248,18 @@ static void gen_exception_insn(DisasContext *s, int offset, int excp, > s->base.is_jmp = DISAS_NORETURN; > } > > +static void gen_exception_bkpt_insn(DisasContext *s, int offset, int syn) Can you use the same Aa64 prototype when applying please? (int syn -> uint32_t syndrome) Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > +{ > + TCGv_i32 tcg_syn; > + > + gen_set_condexec(s); > + gen_set_pc_im(s, s->pc - offset); > + tcg_syn = tcg_const_i32(syn); > + gen_helper_exception_bkpt_insn(cpu_env, tcg_syn); > + tcg_temp_free_i32(tcg_syn); > + s->base.is_jmp = DISAS_NORETURN; > +} > + > /* Force a TB lookup after an instruction that changes the CPU state. */ > static inline void gen_lookup_tb(DisasContext *s) > { > @@ -8774,9 +8786,7 @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) > case 1: > /* bkpt */ > ARCH(5); > - gen_exception_insn(s, 4, EXCP_BKPT, > - syn_aa32_bkpt(imm16, false), > - default_exception_el(s)); > + gen_exception_bkpt_insn(s, 4, syn_aa32_bkpt(imm16, false)); > break; > case 2: > /* Hypervisor call (v7) */ > @@ -11983,8 +11993,7 @@ static void disas_thumb_insn(DisasContext *s, uint32_t insn) > { > int imm8 = extract32(insn, 0, 8); > ARCH(5); > - gen_exception_insn(s, 2, EXCP_BKPT, syn_aa32_bkpt(imm8, true), > - default_exception_el(s)); > + gen_exception_bkpt_insn(s, 2, syn_aa32_bkpt(imm8, true)); > break; > } > > ^ permalink raw reply [flat|nested] 11+ messages in thread
* [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions 2018-03-20 13:41 [Qemu-devel] [PATCH for-2.12 0/4] Fix various BRK/BKPT related bugs Peter Maydell 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK Peter Maydell @ 2018-03-20 13:41 ` Peter Maydell 2018-03-21 22:26 ` Philippe Mathieu-Daudé 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 3/4] target/arm: Set FSR for BKPT, BRK when raising exception Peter Maydell 2018-03-20 13:41 ` [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions Peter Maydell 3 siblings, 1 reply; 11+ messages in thread From: Peter Maydell @ 2018-03-20 13:41 UTC (permalink / raw) To: qemu-arm, qemu-devel; +Cc: patches When a debug exception is taken to AArch32, it appears as a Prefetch Abort, and the Instruction Fault Status Register (IFSR) must be set. The IFSR has two possible formats, depending on whether LPAE is in use. Factor out the code in arm_debug_excp_handler() which picks an FSR value into its own utility function, update it to use arm_fi_to_lfsc() and arm_fi_to_sfsc() rather than hard-coded constants, and use the correct condition to select long or short format. In particular this fixes a bug where we could select the short format because we're at EL0 and the EL1 translation regime is not using LPAE, but then route the debug exception to EL2 because of MDCR_EL2.TDE and hand EL2 the wrong format FSR. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/internals.h | 25 +++++++++++++++++++++++++ target/arm/op_helper.c | 12 ++---------- 2 files changed, 27 insertions(+), 10 deletions(-) diff --git a/target/arm/internals.h b/target/arm/internals.h index 47cc224a46..8ce944b7a0 100644 --- a/target/arm/internals.h +++ b/target/arm/internals.h @@ -763,4 +763,29 @@ static inline bool regime_is_secure(CPUARMState *env, ARMMMUIdx mmu_idx) } } +/* Return the FSR value for a debug exception (watchpoint, hardware + * breakpoint or BKPT insn) targeting the specified exception level. + */ +static inline uint32_t arm_debug_exception_fsr(CPUARMState *env) +{ + ARMMMUFaultInfo fi = { .type = ARMFault_Debug }; + int target_el = arm_debug_target_el(env); + bool using_lpae = false; + + if (target_el == 2 || arm_el_is_aa64(env, target_el)) { + using_lpae = true; + } else { + if (arm_feature(env, ARM_FEATURE_LPAE) && + (env->cp15.tcr_el[target_el].raw_tcr & TTBCR_EAE)) { + using_lpae = true; + } + } + + if (using_lpae) { + return arm_fi_to_lfsc(&fi); + } else { + return arm_fi_to_sfsc(&fi); + } +} + #endif diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c index 4b123d2bd6..75efff9edf 100644 --- a/target/arm/op_helper.c +++ b/target/arm/op_helper.c @@ -1330,11 +1330,7 @@ void arm_debug_excp_handler(CPUState *cs) cs->watchpoint_hit = NULL; - if (extended_addresses_enabled(env)) { - env->exception.fsr = (1 << 9) | 0x22; - } else { - env->exception.fsr = 0x2; - } + env->exception.fsr = arm_debug_exception_fsr(env); env->exception.vaddress = wp_hit->hitaddr; raise_exception(env, EXCP_DATA_ABORT, syn_watchpoint(same_el, 0, wnr), @@ -1354,11 +1350,7 @@ void arm_debug_excp_handler(CPUState *cs) return; } - if (extended_addresses_enabled(env)) { - env->exception.fsr = (1 << 9) | 0x22; - } else { - env->exception.fsr = 0x2; - } + env->exception.fsr = arm_debug_exception_fsr(env); /* FAR is UNKNOWN, so doesn't need setting */ raise_exception(env, EXCP_PREFETCH_ABORT, syn_breakpoint(same_el), -- 2.16.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions Peter Maydell @ 2018-03-21 22:26 ` Philippe Mathieu-Daudé 2018-03-22 10:57 ` Peter Maydell 0 siblings, 1 reply; 11+ messages in thread From: Philippe Mathieu-Daudé @ 2018-03-21 22:26 UTC (permalink / raw) To: Peter Maydell, qemu-arm, qemu-devel; +Cc: patches Hi Peter, This patch might be split in 2. On 03/20/2018 02:41 PM, Peter Maydell wrote: > When a debug exception is taken to AArch32, it appears as a Prefetch > Abort, and the Instruction Fault Status Register (IFSR) must be set. > The IFSR has two possible formats, depending on whether LPAE is in ^ intro > use. Factor out the code in arm_debug_excp_handler() which picks > an FSR value into its own utility function, update it to use > arm_fi_to_lfsc() and arm_fi_to_sfsc() rather than hard-coded constants, ^ part 1 (refactor): Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > and use the correct condition to select long or short format. ^ part 2 (fix) looks correct: Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > > In particular this fixes a bug where we could select the short > format because we're at EL0 and the EL1 translation regime is > not using LPAE, but then route the debug exception to EL2 because > of MDCR_EL2.TDE and hand EL2 the wrong format FSR. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > target/arm/internals.h | 25 +++++++++++++++++++++++++ > target/arm/op_helper.c | 12 ++---------- > 2 files changed, 27 insertions(+), 10 deletions(-) > > diff --git a/target/arm/internals.h b/target/arm/internals.h > index 47cc224a46..8ce944b7a0 100644 > --- a/target/arm/internals.h > +++ b/target/arm/internals.h > @@ -763,4 +763,29 @@ static inline bool regime_is_secure(CPUARMState *env, ARMMMUIdx mmu_idx) > } > } > > +/* Return the FSR value for a debug exception (watchpoint, hardware > + * breakpoint or BKPT insn) targeting the specified exception level. > + */ > +static inline uint32_t arm_debug_exception_fsr(CPUARMState *env) > +{ > + ARMMMUFaultInfo fi = { .type = ARMFault_Debug }; > + int target_el = arm_debug_target_el(env); > + bool using_lpae = false; > + > + if (target_el == 2 || arm_el_is_aa64(env, target_el)) { > + using_lpae = true; > + } else { > + if (arm_feature(env, ARM_FEATURE_LPAE) && > + (env->cp15.tcr_el[target_el].raw_tcr & TTBCR_EAE)) { > + using_lpae = true; > + } > + } This looks pretty similar to regime_using_lpae_format() but for ARMFault_Debug. > + > + if (using_lpae) { > + return arm_fi_to_lfsc(&fi); > + } else { > + return arm_fi_to_sfsc(&fi); > + } > +} > + > #endif > diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c > index 4b123d2bd6..75efff9edf 100644 > --- a/target/arm/op_helper.c > +++ b/target/arm/op_helper.c > @@ -1330,11 +1330,7 @@ void arm_debug_excp_handler(CPUState *cs) > > cs->watchpoint_hit = NULL; > > - if (extended_addresses_enabled(env)) { > - env->exception.fsr = (1 << 9) | 0x22; > - } else { > - env->exception.fsr = 0x2; > - } > + env->exception.fsr = arm_debug_exception_fsr(env); > env->exception.vaddress = wp_hit->hitaddr; > raise_exception(env, EXCP_DATA_ABORT, > syn_watchpoint(same_el, 0, wnr), > @@ -1354,11 +1350,7 @@ void arm_debug_excp_handler(CPUState *cs) > return; > } > > - if (extended_addresses_enabled(env)) { > - env->exception.fsr = (1 << 9) | 0x22; > - } else { > - env->exception.fsr = 0x2; > - } > + env->exception.fsr = arm_debug_exception_fsr(env); > /* FAR is UNKNOWN, so doesn't need setting */ > raise_exception(env, EXCP_PREFETCH_ABORT, > syn_breakpoint(same_el), > ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions 2018-03-21 22:26 ` Philippe Mathieu-Daudé @ 2018-03-22 10:57 ` Peter Maydell 0 siblings, 0 replies; 11+ messages in thread From: Peter Maydell @ 2018-03-22 10:57 UTC (permalink / raw) To: Philippe Mathieu-Daudé; +Cc: qemu-arm, QEMU Developers, patches@linaro.org On 21 March 2018 at 22:26, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote: > Hi Peter, > > This patch might be split in 2. For a +27-10 patch, it doesn't really seem necessary. > On 03/20/2018 02:41 PM, Peter Maydell wrote: >> When a debug exception is taken to AArch32, it appears as a Prefetch >> Abort, and the Instruction Fault Status Register (IFSR) must be set. >> The IFSR has two possible formats, depending on whether LPAE is in > > ^ intro > >> use. Factor out the code in arm_debug_excp_handler() which picks >> an FSR value into its own utility function, update it to use >> arm_fi_to_lfsc() and arm_fi_to_sfsc() rather than hard-coded constants, > > ^ part 1 (refactor): > Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > >> and use the correct condition to select long or short format. > > ^ part 2 (fix) looks correct: > Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > >> >> In particular this fixes a bug where we could select the short >> format because we're at EL0 and the EL1 translation regime is >> not using LPAE, but then route the debug exception to EL2 because >> of MDCR_EL2.TDE and hand EL2 the wrong format FSR. >> >> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> >> --- >> target/arm/internals.h | 25 +++++++++++++++++++++++++ >> target/arm/op_helper.c | 12 ++---------- >> 2 files changed, 27 insertions(+), 10 deletions(-) >> >> diff --git a/target/arm/internals.h b/target/arm/internals.h >> index 47cc224a46..8ce944b7a0 100644 >> --- a/target/arm/internals.h >> +++ b/target/arm/internals.h >> @@ -763,4 +763,29 @@ static inline bool regime_is_secure(CPUARMState *env, ARMMMUIdx mmu_idx) >> } >> } >> >> +/* Return the FSR value for a debug exception (watchpoint, hardware >> + * breakpoint or BKPT insn) targeting the specified exception level. >> + */ >> +static inline uint32_t arm_debug_exception_fsr(CPUARMState *env) >> +{ >> + ARMMMUFaultInfo fi = { .type = ARMFault_Debug }; >> + int target_el = arm_debug_target_el(env); >> + bool using_lpae = false; >> + >> + if (target_el == 2 || arm_el_is_aa64(env, target_el)) { >> + using_lpae = true; >> + } else { >> + if (arm_feature(env, ARM_FEATURE_LPAE) && >> + (env->cp15.tcr_el[target_el].raw_tcr & TTBCR_EAE)) { >> + using_lpae = true; >> + } >> + } > > This looks pretty similar to regime_using_lpae_format() but for > ARMFault_Debug. Yeah, it's basically similar logic. I'm a bit confused overall -- are you giving a reviewed-by for this patch, or do you want changes? thanks -- PMM ^ permalink raw reply [flat|nested] 11+ messages in thread
* [Qemu-arm] [PATCH for-2.12 3/4] target/arm: Set FSR for BKPT, BRK when raising exception 2018-03-20 13:41 [Qemu-devel] [PATCH for-2.12 0/4] Fix various BRK/BKPT related bugs Peter Maydell 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK Peter Maydell 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions Peter Maydell @ 2018-03-20 13:41 ` Peter Maydell 2018-03-22 8:23 ` [Qemu-devel] " Philippe Mathieu-Daudé 2018-03-20 13:41 ` [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions Peter Maydell 3 siblings, 1 reply; 11+ messages in thread From: Peter Maydell @ 2018-03-20 13:41 UTC (permalink / raw) To: qemu-arm, qemu-devel; +Cc: patches Now that we have a helper function specifically for the BRK and BKPT instructions, we can set the exception.fsr there rather than in arm_cpu_do_interrupt_aarch32(). This allows us to use our new arm_debug_exception_fsr() helper. In particular this fixes a bug where we were hardcoding the short-form IFSR value, which is wrong if the target exception level has LPAE enabled. Fixes: https://bugs.launchpad.net/qemu/+bug/1756927 Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 1 - target/arm/op_helper.c | 2 ++ 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index 09893e3f72..dcb8476d9e 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -7910,7 +7910,6 @@ static void arm_cpu_do_interrupt_aarch32(CPUState *cs) offset = 0; break; case EXCP_BKPT: - env->exception.fsr = 2; /* Fall through to prefetch abort. */ case EXCP_PREFETCH_ABORT: A32_BANKED_CURRENT_REG_SET(env, ifsr, env->exception.fsr); diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c index 75efff9edf..8e1e521193 100644 --- a/target/arm/op_helper.c +++ b/target/arm/op_helper.c @@ -488,6 +488,8 @@ void HELPER(exception_with_syndrome)(CPUARMState *env, uint32_t excp, */ void HELPER(exception_bkpt_insn)(CPUARMState *env, uint32_t syndrome) { + /* FSR will only be used if the debug target EL is AArch32. */ + env->exception.fsr = arm_debug_exception_fsr(env); raise_exception(env, EXCP_BKPT, syndrome, arm_debug_target_el(env)); } -- 2.16.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [Qemu-arm] [PATCH for-2.12 3/4] target/arm: Set FSR for BKPT, BRK when raising exception 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 3/4] target/arm: Set FSR for BKPT, BRK when raising exception Peter Maydell @ 2018-03-22 8:23 ` Philippe Mathieu-Daudé 0 siblings, 0 replies; 11+ messages in thread From: Philippe Mathieu-Daudé @ 2018-03-22 8:23 UTC (permalink / raw) To: Peter Maydell, qemu-arm, qemu-devel; +Cc: patches On 03/20/2018 10:41 AM, Peter Maydell wrote: > Now that we have a helper function specifically for the BRK and > BKPT instructions, we can set the exception.fsr there rather > than in arm_cpu_do_interrupt_aarch32(). This allows us to > use our new arm_debug_exception_fsr() helper. > > In particular this fixes a bug where we were hardcoding the > short-form IFSR value, which is wrong if the target exception > level has LPAE enabled. > > Fixes: https://bugs.launchpad.net/qemu/+bug/1756927 > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > --- > target/arm/helper.c | 1 - > target/arm/op_helper.c | 2 ++ > 2 files changed, 2 insertions(+), 1 deletion(-) > > diff --git a/target/arm/helper.c b/target/arm/helper.c > index 09893e3f72..dcb8476d9e 100644 > --- a/target/arm/helper.c > +++ b/target/arm/helper.c > @@ -7910,7 +7910,6 @@ static void arm_cpu_do_interrupt_aarch32(CPUState *cs) > offset = 0; > break; > case EXCP_BKPT: > - env->exception.fsr = 2; > /* Fall through to prefetch abort. */ > case EXCP_PREFETCH_ABORT: > A32_BANKED_CURRENT_REG_SET(env, ifsr, env->exception.fsr); > diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c > index 75efff9edf..8e1e521193 100644 > --- a/target/arm/op_helper.c > +++ b/target/arm/op_helper.c > @@ -488,6 +488,8 @@ void HELPER(exception_with_syndrome)(CPUARMState *env, uint32_t excp, > */ > void HELPER(exception_bkpt_insn)(CPUARMState *env, uint32_t syndrome) > { > + /* FSR will only be used if the debug target EL is AArch32. */ > + env->exception.fsr = arm_debug_exception_fsr(env); > raise_exception(env, EXCP_BKPT, syndrome, arm_debug_target_el(env)); > } > > ^ permalink raw reply [flat|nested] 11+ messages in thread
* [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions 2018-03-20 13:41 [Qemu-devel] [PATCH for-2.12 0/4] Fix various BRK/BKPT related bugs Peter Maydell ` (2 preceding siblings ...) 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 3/4] target/arm: Set FSR for BKPT, BRK when raising exception Peter Maydell @ 2018-03-20 13:41 ` Peter Maydell 2018-03-21 22:29 ` [Qemu-arm] " Philippe Mathieu-Daudé 3 siblings, 1 reply; 11+ messages in thread From: Peter Maydell @ 2018-03-20 13:41 UTC (permalink / raw) To: qemu-arm, qemu-devel; +Cc: patches For debug exceptions due to breakpoints or the BKPT instruction which are taken to AArch32, the Fault Address Register is architecturally UNKNOWN. We were using that as license to simply not set env->exception.vaddress, but this isn't correct, because it will expose to the guest whatever old value was in that field when arm_cpu_do_interrupt_aarch32() writes it to the guest IFSR. That old value might be a FAR for a previous guest EL2 or secure exception, in which case we shouldn't show it to an EL1 or non-secure exception handler. It might also be a non-deterministic value, which is bad for record-and-replay. Clear env->exception.vaddress before taking breakpoint debug exceptions, to avoid this minor information leak. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/op_helper.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c index 8e1e521193..a266cc0116 100644 --- a/target/arm/op_helper.c +++ b/target/arm/op_helper.c @@ -490,6 +490,11 @@ void HELPER(exception_bkpt_insn)(CPUARMState *env, uint32_t syndrome) { /* FSR will only be used if the debug target EL is AArch32. */ env->exception.fsr = arm_debug_exception_fsr(env); + /* FAR is UNKNOWN: clear vaddress to avoid potentially exposing + * values to the guest that it shouldn't be able to see at its + * exception/security level. + */ + env->exception.vaddress = 0; raise_exception(env, EXCP_BKPT, syndrome, arm_debug_target_el(env)); } @@ -1353,7 +1358,11 @@ void arm_debug_excp_handler(CPUState *cs) } env->exception.fsr = arm_debug_exception_fsr(env); - /* FAR is UNKNOWN, so doesn't need setting */ + /* FAR is UNKNOWN: clear vaddress to avoid potentially exposing + * values to the guest that it shouldn't be able to see at its + * exception/security level. + */ + env->exception.vaddress = 0; raise_exception(env, EXCP_PREFETCH_ABORT, syn_breakpoint(same_el), arm_debug_target_el(env)); -- 2.16.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-arm] [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions 2018-03-20 13:41 ` [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions Peter Maydell @ 2018-03-21 22:29 ` Philippe Mathieu-Daudé 2018-03-22 10:59 ` Peter Maydell 0 siblings, 1 reply; 11+ messages in thread From: Philippe Mathieu-Daudé @ 2018-03-21 22:29 UTC (permalink / raw) To: Peter Maydell, qemu-arm, qemu-devel; +Cc: patches On 03/20/2018 02:41 PM, Peter Maydell wrote: > For debug exceptions due to breakpoints or the BKPT instruction which > are taken to AArch32, the Fault Address Register is architecturally > UNKNOWN. We were using that as license to simply not set > env->exception.vaddress, but this isn't correct, because it will > expose to the guest whatever old value was in that field when > arm_cpu_do_interrupt_aarch32() writes it to the guest IFSR. That old > value might be a FAR for a previous guest EL2 or secure exception, in > which case we shouldn't show it to an EL1 or non-secure exception > handler. It might also be a non-deterministic value, which is bad > for record-and-replay. > > Clear env->exception.vaddress before taking breakpoint debug > exceptions, to avoid this minor information leak. So this series is worth Cc'ing qemu-stable...? > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> > --- > target/arm/op_helper.c | 11 ++++++++++- > 1 file changed, 10 insertions(+), 1 deletion(-) > > diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c > index 8e1e521193..a266cc0116 100644 > --- a/target/arm/op_helper.c > +++ b/target/arm/op_helper.c > @@ -490,6 +490,11 @@ void HELPER(exception_bkpt_insn)(CPUARMState *env, uint32_t syndrome) > { > /* FSR will only be used if the debug target EL is AArch32. */ > env->exception.fsr = arm_debug_exception_fsr(env); > + /* FAR is UNKNOWN: clear vaddress to avoid potentially exposing > + * values to the guest that it shouldn't be able to see at its > + * exception/security level. > + */ > + env->exception.vaddress = 0; > raise_exception(env, EXCP_BKPT, syndrome, arm_debug_target_el(env)); > } > > @@ -1353,7 +1358,11 @@ void arm_debug_excp_handler(CPUState *cs) > } > > env->exception.fsr = arm_debug_exception_fsr(env); > - /* FAR is UNKNOWN, so doesn't need setting */ > + /* FAR is UNKNOWN: clear vaddress to avoid potentially exposing > + * values to the guest that it shouldn't be able to see at its > + * exception/security level. > + */ > + env->exception.vaddress = 0; > raise_exception(env, EXCP_PREFETCH_ABORT, > syn_breakpoint(same_el), > arm_debug_target_el(env)); > ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions 2018-03-21 22:29 ` [Qemu-arm] " Philippe Mathieu-Daudé @ 2018-03-22 10:59 ` Peter Maydell 0 siblings, 0 replies; 11+ messages in thread From: Peter Maydell @ 2018-03-22 10:59 UTC (permalink / raw) To: Philippe Mathieu-Daudé; +Cc: qemu-arm, QEMU Developers, patches@linaro.org On 21 March 2018 at 22:29, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote: > On 03/20/2018 02:41 PM, Peter Maydell wrote: >> For debug exceptions due to breakpoints or the BKPT instruction which >> are taken to AArch32, the Fault Address Register is architecturally >> UNKNOWN. We were using that as license to simply not set >> env->exception.vaddress, but this isn't correct, because it will >> expose to the guest whatever old value was in that field when >> arm_cpu_do_interrupt_aarch32() writes it to the guest IFSR. That old >> value might be a FAR for a previous guest EL2 or secure exception, in >> which case we shouldn't show it to an EL1 or non-secure exception >> handler. It might also be a non-deterministic value, which is bad >> for record-and-replay. >> >> Clear env->exception.vaddress before taking breakpoint debug >> exceptions, to avoid this minor information leak. > > So this series is worth Cc'ing qemu-stable...? I don't think so. This isn't a regression -- we've behaved this way ever since we introduced LPAE support, five years or so ago. In particular Linux doesn't care about it. thanks -- PMM ^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2018-03-22 11:00 UTC | newest] Thread overview: 11+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2018-03-20 13:41 [Qemu-devel] [PATCH for-2.12 0/4] Fix various BRK/BKPT related bugs Peter Maydell 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 1/4] target/arm: Honour MDCR_EL2.TDE when routing exceptions due to BKPT/BRK Peter Maydell 2018-03-22 8:22 ` Philippe Mathieu-Daudé 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 2/4] target/arm: Factor out code to calculate FSR for debug exceptions Peter Maydell 2018-03-21 22:26 ` Philippe Mathieu-Daudé 2018-03-22 10:57 ` Peter Maydell 2018-03-20 13:41 ` [Qemu-arm] [PATCH for-2.12 3/4] target/arm: Set FSR for BKPT, BRK when raising exception Peter Maydell 2018-03-22 8:23 ` [Qemu-devel] " Philippe Mathieu-Daudé 2018-03-20 13:41 ` [Qemu-devel] [PATCH for-2.12 4/4] target/arm: Always set FAR to a known unknown value for debug exceptions Peter Maydell 2018-03-21 22:29 ` [Qemu-arm] " Philippe Mathieu-Daudé 2018-03-22 10:59 ` Peter Maydell
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).