| Message ID | 761ce615438312ca5d37fb8829a0ba903375392f.1700844359.git.simone.ballarin@bugseng.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | xen: address violations of MISRA C:2012 Rule 13.1 | expand |
On 24.11.2023 18:29, Simone Ballarin wrote: > Rule 13.1: Initializer lists shall not contain persistent side effects > > Effects caused by debug/logging macros and functions (like ASSERT, __bad_atomic_size, > LOG, etc ...) that crash execution or produce logs are not dangerous in initializer > lists. The evaluation order in abnormal conditions is not relevant. Evaluation order > of logging effects is always safe. > > This patch deviates violations using SAF commits caused by debug/logging macros and > functions. > > Asm volatile statements in initializer lists that do not perform any persistent side > effect are safe: this patch deviates violations caused by uses of the current macro > (that contains an asm volatile) in initializer lists. > > No functional changes. > > Signed-off-by: Simone Ballarin <simone.ballarin@bugseng.com> > > --- > Changes in v2: > New patch based on the discussion for "xen/arm: address violations of MISRA C:2012 Rule 13.1". > --- > docs/misra/safe.json | 16 ++++++++++++++++ > xen/arch/arm/device.c | 1 + > xen/arch/arm/guestcopy.c | 4 ++++ > xen/arch/x86/hvm/hvm.c | 1 + > xen/common/sched/core.c | 3 +++ The latter two don't really fit the title prefix. > --- a/docs/misra/safe.json > +++ b/docs/misra/safe.json > @@ -28,6 +28,22 @@ > }, > { > "id": "SAF-3-safe", > + "analyser": { > + "eclair": "MC3R1.R13.1" > + }, > + "name": "MC3R1.R13.1: effects for debugging and logging", > + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." > + }, I'm wary of this statement. Order may not matter much anymore _after_ an abnormal condition was encountered, but in the course of determining whether an abnormal condition might have been reached it may very well still matter. > + { > + "id": "SAF-4-safe", > + "analyser": { > + "eclair": "MC3R1.R13.1" > + }, > + "name": "MC3R1.R13.1: volatile asm statements that do not perform any persistent side effect", > + "text": "Volatile asm statements in an initializer list if do not perform persistent side effects are safe." Since each respective comment ought to affect just a single asm(), I think this wants writing in singular. I further don't think it is useful for "text" to largely restate what "name" already says. > --- a/xen/arch/arm/device.c > +++ b/xen/arch/arm/device.c > @@ -331,6 +331,7 @@ int handle_device(struct domain *d, struct dt_device_node *dev, p2m_type_t p2mt, > .p2mt = p2mt, > .skip_mapping = !own_device || > (is_pci_passthrough_enabled() && > + /* SAF-3-safe effects for debugging/logging reasons are safe */ > (device_get_class(dev) == DEVICE_PCI_HOSTBRIDGE)), What's the debugging / logging reason on the commented line? > --- a/xen/arch/arm/guestcopy.c > +++ b/xen/arch/arm/guestcopy.c > @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, > unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) > { > return copy_guest((void *)from, (vaddr_t)to, len, > + /* SAF-4-safe No persistent side effects */ > GVA_INFO(current), COPY_to_guest | COPY_linear); > } > > unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, > unsigned int len) > { > + /* SAF-4-safe No persistent side effects */ > return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), > COPY_to_guest | COPY_flush_dcache | COPY_linear); > } > > unsigned long raw_clear_guest(void *to, unsigned int len) > { > + /* SAF-4-safe No persistent side effects */ > return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), > COPY_to_guest | COPY_linear); > } > @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) > unsigned long raw_copy_from_guest(void *to, const void __user *from, > unsigned int len) > { > + /* SAF-4-safe No persistent side effects */ > return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), > COPY_from_guest | COPY_linear); > } I can only guess that in all four of these it's the use of "current" which requires the comment. Yet imo that either needs making explicit, or such a comment shouldn't go on use sites of "current", but on its definition site. > --- a/xen/common/sched/core.c > +++ b/xen/common/sched/core.c > @@ -1517,6 +1517,7 @@ long vcpu_yield(void) > > SCHED_STAT_CRANK(vcpu_yield); > > + /* SAF-4-safe No persistent side effects */ > TRACE_2D(TRC_SCHED_YIELD, current->domain->domain_id, current->vcpu_id); > raise_softirq(SCHEDULE_SOFTIRQ); > return 0; > @@ -1895,6 +1896,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) > if ( copy_from_guest(&sched_shutdown, arg, 1) ) > break; > > + /* SAF-4-safe No persistent side effects */ > TRACE_3D(TRC_SCHED_SHUTDOWN, > current->domain->domain_id, current->vcpu_id, > sched_shutdown.reason); > @@ -1912,6 +1914,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) > if ( copy_from_guest(&sched_shutdown, arg, 1) ) > break; > > + /* SAF-4-safe No persistent side effects */ > TRACE_3D(TRC_SCHED_SHUTDOWN_CODE, > d->domain_id, current->vcpu_id, sched_shutdown.reason); > In at least the former two of these cases pulling out "current" into a local variable "curr" would likely eliminate the violation and at the same time improve code a little. Jan
On 27/11/23 11:46, Jan Beulich wrote: > On 24.11.2023 18:29, Simone Ballarin wrote: >> Rule 13.1: Initializer lists shall not contain persistent side effects >> >> Effects caused by debug/logging macros and functions (like ASSERT, __bad_atomic_size, >> LOG, etc ...) that crash execution or produce logs are not dangerous in initializer >> lists. The evaluation order in abnormal conditions is not relevant. Evaluation order >> of logging effects is always safe. >> >> This patch deviates violations using SAF commits caused by debug/logging macros and >> functions. >> >> Asm volatile statements in initializer lists that do not perform any persistent side >> effect are safe: this patch deviates violations caused by uses of the current macro >> (that contains an asm volatile) in initializer lists. >> >> No functional changes. >> >> Signed-off-by: Simone Ballarin <simone.ballarin@bugseng.com> >> >> --- >> Changes in v2: >> New patch based on the discussion for "xen/arm: address violations of MISRA C:2012 Rule 13.1". >> --- >> docs/misra/safe.json | 16 ++++++++++++++++ >> xen/arch/arm/device.c | 1 + >> xen/arch/arm/guestcopy.c | 4 ++++ >> xen/arch/x86/hvm/hvm.c | 1 + >> xen/common/sched/core.c | 3 +++ > > The latter two don't really fit the title prefix. > >> --- a/docs/misra/safe.json >> +++ b/docs/misra/safe.json >> @@ -28,6 +28,22 @@ >> }, >> { >> "id": "SAF-3-safe", >> + "analyser": { >> + "eclair": "MC3R1.R13.1" >> + }, >> + "name": "MC3R1.R13.1: effects for debugging and logging", >> + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." >> + }, > > I'm wary of this statement. Order may not matter much anymore _after_ an > abnormal condition was encountered, but in the course of determining whether > an abnormal condition might have been reached it may very well still matter. > Do you object to the deviation in general? Or just to the wording? >> + { >> + "id": "SAF-4-safe", >> + "analyser": { >> + "eclair": "MC3R1.R13.1" >> + }, >> + "name": "MC3R1.R13.1: volatile asm statements that do not perform any persistent side effect", >> + "text": "Volatile asm statements in an initializer list if do not perform persistent side effects are safe." > > Since each respective comment ought to affect just a single asm(), I think > this wants writing in singular. I further don't think it is useful for > "text" to largely restate what "name" already says. > Ok. >> --- a/xen/arch/arm/device.c >> +++ b/xen/arch/arm/device.c >> @@ -331,6 +331,7 @@ int handle_device(struct domain *d, struct dt_device_node *dev, p2m_type_t p2mt, >> .p2mt = p2mt, >> .skip_mapping = !own_device || >> (is_pci_passthrough_enabled() && >> + /* SAF-3-safe effects for debugging/logging reasons are safe */ >> (device_get_class(dev) == DEVICE_PCI_HOSTBRIDGE)), > > What's the debugging / logging reason on the commented line? The "ASSERT(dev != NULL)" in its body. > >> --- a/xen/arch/arm/guestcopy.c >> +++ b/xen/arch/arm/guestcopy.c >> @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, >> unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) >> { >> return copy_guest((void *)from, (vaddr_t)to, len, >> + /* SAF-4-safe No persistent side effects */ >> GVA_INFO(current), COPY_to_guest | COPY_linear); >> } >> >> unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, >> unsigned int len) >> { >> + /* SAF-4-safe No persistent side effects */ >> return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), >> COPY_to_guest | COPY_flush_dcache | COPY_linear); >> } >> >> unsigned long raw_clear_guest(void *to, unsigned int len) >> { >> + /* SAF-4-safe No persistent side effects */ >> return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), >> COPY_to_guest | COPY_linear); >> } >> @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) >> unsigned long raw_copy_from_guest(void *to, const void __user *from, >> unsigned int len) >> { >> + /* SAF-4-safe No persistent side effects */ >> return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), >> COPY_from_guest | COPY_linear); >> } > > I can only guess that in all four of these it's the use of "current" which > requires the comment. Yet imo that either needs making explicit, or such a > comment shouldn't go on use sites of "current", but on its definition site. > "current" does not contain any violation of R13.1. Its expansion produces a side-effect, but this is not a problem in itself. The real problem is that GVA_INFO expands it in an initializer list: #define GVA_INFO(vcpu) ((copy_info_t) { .gva = { vcpu } }) GVA_INFO(current) is the real piece of code that requires to be deviated. >> --- a/xen/common/sched/core.c >> +++ b/xen/common/sched/core.c >> @@ -1517,6 +1517,7 @@ long vcpu_yield(void) >> >> SCHED_STAT_CRANK(vcpu_yield); >> >> + /* SAF-4-safe No persistent side effects */ >> TRACE_2D(TRC_SCHED_YIELD, current->domain->domain_id, current->vcpu_id); >> raise_softirq(SCHEDULE_SOFTIRQ); >> return 0; >> @@ -1895,6 +1896,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) >> if ( copy_from_guest(&sched_shutdown, arg, 1) ) >> break; >> >> + /* SAF-4-safe No persistent side effects */ >> TRACE_3D(TRC_SCHED_SHUTDOWN, >> current->domain->domain_id, current->vcpu_id, >> sched_shutdown.reason); >> @@ -1912,6 +1914,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) >> if ( copy_from_guest(&sched_shutdown, arg, 1) ) >> break; >> >> + /* SAF-4-safe No persistent side effects */ >> TRACE_3D(TRC_SCHED_SHUTDOWN_CODE, >> d->domain_id, current->vcpu_id, sched_shutdown.reason); >> > > In at least the former two of these cases pulling out "current" into a local > variable "curr" would likely eliminate the violation and at the same time > improve code a little. > Yes, I agree. > Jan >
On 27.11.2023 15:35, Simone Ballarin wrote: > On 27/11/23 11:46, Jan Beulich wrote: >> On 24.11.2023 18:29, Simone Ballarin wrote: >>> --- a/docs/misra/safe.json >>> +++ b/docs/misra/safe.json >>> @@ -28,6 +28,22 @@ >>> }, >>> { >>> "id": "SAF-3-safe", >>> + "analyser": { >>> + "eclair": "MC3R1.R13.1" >>> + }, >>> + "name": "MC3R1.R13.1: effects for debugging and logging", >>> + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." >>> + }, >> >> I'm wary of this statement. Order may not matter much anymore _after_ an >> abnormal condition was encountered, but in the course of determining whether >> an abnormal condition might have been reached it may very well still matter. > > Do you object to the deviation in general? Or just to the wording? Primarily the wording. Yet the need to adjust the wording also hints at there needing to be care when actually making use of this deviation. Which it turn I'm not convinced is in the spirit of Misra. >>> --- a/xen/arch/arm/device.c >>> +++ b/xen/arch/arm/device.c >>> @@ -331,6 +331,7 @@ int handle_device(struct domain *d, struct dt_device_node *dev, p2m_type_t p2mt, >>> .p2mt = p2mt, >>> .skip_mapping = !own_device || >>> (is_pci_passthrough_enabled() && >>> + /* SAF-3-safe effects for debugging/logging reasons are safe */ >>> (device_get_class(dev) == DEVICE_PCI_HOSTBRIDGE)), >> >> What's the debugging / logging reason on the commented line? > > The "ASSERT(dev != NULL)" in its body. "it" == device_get_class() I assume? Imo to unobvious to deal with like this. >>> --- a/xen/arch/arm/guestcopy.c >>> +++ b/xen/arch/arm/guestcopy.c >>> @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, >>> unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) >>> { >>> return copy_guest((void *)from, (vaddr_t)to, len, >>> + /* SAF-4-safe No persistent side effects */ >>> GVA_INFO(current), COPY_to_guest | COPY_linear); >>> } >>> >>> unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, >>> unsigned int len) >>> { >>> + /* SAF-4-safe No persistent side effects */ >>> return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), >>> COPY_to_guest | COPY_flush_dcache | COPY_linear); >>> } >>> >>> unsigned long raw_clear_guest(void *to, unsigned int len) >>> { >>> + /* SAF-4-safe No persistent side effects */ >>> return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), >>> COPY_to_guest | COPY_linear); >>> } >>> @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) >>> unsigned long raw_copy_from_guest(void *to, const void __user *from, >>> unsigned int len) >>> { >>> + /* SAF-4-safe No persistent side effects */ >>> return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), >>> COPY_from_guest | COPY_linear); >>> } >> >> I can only guess that in all four of these it's the use of "current" which >> requires the comment. Yet imo that either needs making explicit, or such a >> comment shouldn't go on use sites of "current", but on its definition site. >> > > "current" does not contain any violation of R13.1. Its expansion > produces a side-effect, but this is not a problem in itself. The real > problem is that GVA_INFO expands it in an initializer list: > #define GVA_INFO(vcpu) ((copy_info_t) { .gva = { vcpu } }) But an initializer list doesn't itself constitute a side effect, does it? > GVA_INFO(current) is the real piece of code that requires to be > deviated. In which case I think this such want spelling return copy_guest(to, (vaddr_t)from, len, /* SAF-4-safe No persistent side effects */ GVA_INFO(current), COPY_from_guest | COPY_linear); or some such. Not the least because we want to make sure a deviation does not have wider than necessary scope (which when formatted as above is, aiui, always the following source line). Jan
On 27/11/23 16:09, Jan Beulich wrote: > On 27.11.2023 15:35, Simone Ballarin wrote: >> On 27/11/23 11:46, Jan Beulich wrote: >>> On 24.11.2023 18:29, Simone Ballarin wrote: >>>> --- a/docs/misra/safe.json >>>> +++ b/docs/misra/safe.json >>>> @@ -28,6 +28,22 @@ >>>> }, >>>> { >>>> "id": "SAF-3-safe", >>>> + "analyser": { >>>> + "eclair": "MC3R1.R13.1" >>>> + }, >>>> + "name": "MC3R1.R13.1: effects for debugging and logging", >>>> + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." >>>> + }, >>> >>> I'm wary of this statement. Order may not matter much anymore _after_ an >>> abnormal condition was encountered, but in the course of determining whether >>> an abnormal condition might have been reached it may very well still matter. >> >> Do you object to the deviation in general? Or just to the wording? > > Primarily the wording. Yet the need to adjust the wording also hints at there > needing to be care when actually making use of this deviation. Which it turn > I'm not convinced is in the spirit of Misra The rule is really strict, but imho the only real dangerous situation is when an entry performs a persistent side effect that can change the behavior of another entry. I.e.: int y = 0; int x[2] = { y=1, /* first element will be always 1 */ y /* second element can be either 0 or 1 */ }; Above we have a dependency between the first entry and the second. Now let's consider logging effects: #define LOG(x) printf("LOG: %s", x); int x[2] = { ({ LOG("1"); 1; }), ({ LOG("2"); 2; }) }; Here the execution can print: "LOG: 1LOG: 2" or "LOG: 2LOG: 1". Do we agree that the evaluation order of prints caused by logging functions/macros do not normally cause dependencies between the entries? The execution is still non-deterministic, but does it really matter?. In the case of function that crash execution, no dependencies can exist since no further entries will be evaluated. In conclusion, I propose the following rewording: "text": "Effects that crash execution or produce logs are allowed in initializer lists. Logging effects do not affect the evaluation of subsequent entries. Crash effects are allowed as they represent the end of the execution. > >>>> --- a/xen/arch/arm/device.c >>>> +++ b/xen/arch/arm/device.c >>>> @@ -331,6 +331,7 @@ int handle_device(struct domain *d, struct dt_device_node *dev, p2m_type_t p2mt, >>>> .p2mt = p2mt, >>>> .skip_mapping = !own_device || >>>> (is_pci_passthrough_enabled() && >>>> + /* SAF-3-safe effects for debugging/logging reasons are safe */ >>>> (device_get_class(dev) == DEVICE_PCI_HOSTBRIDGE)), >>> >>> What's the debugging / logging reason on the commented line? >> >> The "ASSERT(dev != NULL)" in its body. > > "it" == device_get_class() I assume? Imo to unobvious to deal with like this. > Yes, Julien suggested to simply remove the ASSERT as it isn't really useful. I forgot his suggestion. In v3 I will remove it as suggested. >>>> --- a/xen/arch/arm/guestcopy.c >>>> +++ b/xen/arch/arm/guestcopy.c >>>> @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, >>>> unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) >>>> { >>>> return copy_guest((void *)from, (vaddr_t)to, len, >>>> + /* SAF-4-safe No persistent side effects */ >>>> GVA_INFO(current), COPY_to_guest | COPY_linear); >>>> } >>>> >>>> unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, >>>> unsigned int len) >>>> { >>>> + /* SAF-4-safe No persistent side effects */ >>>> return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), >>>> COPY_to_guest | COPY_flush_dcache | COPY_linear); >>>> } >>>> >>>> unsigned long raw_clear_guest(void *to, unsigned int len) >>>> { >>>> + /* SAF-4-safe No persistent side effects */ >>>> return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), >>>> COPY_to_guest | COPY_linear); >>>> } >>>> @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) >>>> unsigned long raw_copy_from_guest(void *to, const void __user *from, >>>> unsigned int len) >>>> { >>>> + /* SAF-4-safe No persistent side effects */ >>>> return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), >>>> COPY_from_guest | COPY_linear); >>>> } >>> >>> I can only guess that in all four of these it's the use of "current" which >>> requires the comment. Yet imo that either needs making explicit, or such a >>> comment shouldn't go on use sites of "current", but on its definition site. >>> >> >> "current" does not contain any violation of R13.1. Its expansion >> produces a side-effect, but this is not a problem in itself. The real >> problem is that GVA_INFO expands it in an initializer list: >> #define GVA_INFO(vcpu) ((copy_info_t) { .gva = { vcpu } }) > > But an initializer list doesn't itself constitute a side effect, does it? > The side effect should be inside the initializer list. { .gva = 1 } is not a violation. >> GVA_INFO(current) is the real piece of code that requires to be >> deviated. > > In which case I think this such want spelling > > return copy_guest(to, (vaddr_t)from, len, > /* SAF-4-safe No persistent side effects */ > GVA_INFO(current), > COPY_from_guest | COPY_linear); > > or some such. Not the least because we want to make sure a deviation does > not have wider than necessary scope (which when formatted as above is, > aiui, always the following source line). > Yes, I agree. > Jan
On 27.11.2023 18:34, Simone Ballarin wrote: > On 27/11/23 16:09, Jan Beulich wrote: >> On 27.11.2023 15:35, Simone Ballarin wrote: >>> On 27/11/23 11:46, Jan Beulich wrote: >>>> On 24.11.2023 18:29, Simone Ballarin wrote: >>>>> --- a/docs/misra/safe.json >>>>> +++ b/docs/misra/safe.json >>>>> @@ -28,6 +28,22 @@ >>>>> }, >>>>> { >>>>> "id": "SAF-3-safe", >>>>> + "analyser": { >>>>> + "eclair": "MC3R1.R13.1" >>>>> + }, >>>>> + "name": "MC3R1.R13.1: effects for debugging and logging", >>>>> + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." >>>>> + }, >>>> >>>> I'm wary of this statement. Order may not matter much anymore _after_ an >>>> abnormal condition was encountered, but in the course of determining whether >>>> an abnormal condition might have been reached it may very well still matter. >>> >>> Do you object to the deviation in general? Or just to the wording? >> >> Primarily the wording. Yet the need to adjust the wording also hints at there >> needing to be care when actually making use of this deviation. Which it turn >> I'm not convinced is in the spirit of Misra > > The rule is really strict, but imho the only real dangerous situation is > when an entry performs a persistent side effect that can change the > behavior of another entry. I.e.: > > int y = 0; > int x[2] = > { > y=1, /* first element will be always 1 */ > y /* second element can be either 0 or 1 */ > }; > > Above we have a dependency between the first entry and the second. > > Now let's consider logging effects: > > #define LOG(x) printf("LOG: %s", x); > > int x[2] = > { > ({ LOG("1"); 1; }), > ({ LOG("2"); 2; }) > }; > > > Here the execution can print: > "LOG: 1LOG: 2" or > "LOG: 2LOG: 1". > > Do we agree that the evaluation order of prints caused by logging > functions/macros do not normally cause dependencies between the > entries? The execution is still non-deterministic, but does it really > matter?. > > In the case of function that crash execution, no dependencies can exist > since no further entries will be evaluated. > > In conclusion, I propose the following rewording: > > "text": "Effects that crash execution or produce logs are allowed in > initializer lists. Logging effects do not affect the evaluation of > subsequent entries. Crash effects are allowed as they represent the > end of the execution. Let's assume we have a BUG_ON() (as the "crash effect") the condition of which depends on where in the sequence of operations it actually executes, i.e. (potentially) dependent upon another part of the initializer. I hope we agree that's not something that should be deviated? Yet even the re- worded statement would - according to my reading - permit doing so. I guess I should try to remember to bring this up on this afternoon's call. >>>>> --- a/xen/arch/arm/guestcopy.c >>>>> +++ b/xen/arch/arm/guestcopy.c >>>>> @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, >>>>> unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) >>>>> { >>>>> return copy_guest((void *)from, (vaddr_t)to, len, >>>>> + /* SAF-4-safe No persistent side effects */ >>>>> GVA_INFO(current), COPY_to_guest | COPY_linear); >>>>> } >>>>> >>>>> unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, >>>>> unsigned int len) >>>>> { >>>>> + /* SAF-4-safe No persistent side effects */ >>>>> return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), >>>>> COPY_to_guest | COPY_flush_dcache | COPY_linear); >>>>> } >>>>> >>>>> unsigned long raw_clear_guest(void *to, unsigned int len) >>>>> { >>>>> + /* SAF-4-safe No persistent side effects */ >>>>> return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), >>>>> COPY_to_guest | COPY_linear); >>>>> } >>>>> @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) >>>>> unsigned long raw_copy_from_guest(void *to, const void __user *from, >>>>> unsigned int len) >>>>> { >>>>> + /* SAF-4-safe No persistent side effects */ >>>>> return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), >>>>> COPY_from_guest | COPY_linear); >>>>> } >>>> >>>> I can only guess that in all four of these it's the use of "current" which >>>> requires the comment. Yet imo that either needs making explicit, or such a >>>> comment shouldn't go on use sites of "current", but on its definition site. >>>> >>> >>> "current" does not contain any violation of R13.1. Its expansion >>> produces a side-effect, but this is not a problem in itself. The real >>> problem is that GVA_INFO expands it in an initializer list: >>> #define GVA_INFO(vcpu) ((copy_info_t) { .gva = { vcpu } }) >> >> But an initializer list doesn't itself constitute a side effect, does it? > > The side effect should be inside the initializer list. { .gva = 1 } is > not a violation. I'm afraid I don't see what would be constituting a violation here. Jan
On 28/11/23 09:42, Jan Beulich wrote: > On 27.11.2023 18:34, Simone Ballarin wrote: >> On 27/11/23 16:09, Jan Beulich wrote: >>> On 27.11.2023 15:35, Simone Ballarin wrote: >>>> On 27/11/23 11:46, Jan Beulich wrote: >>>>> On 24.11.2023 18:29, Simone Ballarin wrote: >>>>>> --- a/docs/misra/safe.json >>>>>> +++ b/docs/misra/safe.json >>>>>> @@ -28,6 +28,22 @@ >>>>>> }, >>>>>> { >>>>>> "id": "SAF-3-safe", >>>>>> + "analyser": { >>>>>> + "eclair": "MC3R1.R13.1" >>>>>> + }, >>>>>> + "name": "MC3R1.R13.1: effects for debugging and logging", >>>>>> + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." >>>>>> + }, >>>>> >>>>> I'm wary of this statement. Order may not matter much anymore _after_ an >>>>> abnormal condition was encountered, but in the course of determining whether >>>>> an abnormal condition might have been reached it may very well still matter. >>>> >>>> Do you object to the deviation in general? Or just to the wording? >>> >>> Primarily the wording. Yet the need to adjust the wording also hints at there >>> needing to be care when actually making use of this deviation. Which it turn >>> I'm not convinced is in the spirit of Misra >> >> The rule is really strict, but imho the only real dangerous situation is >> when an entry performs a persistent side effect that can change the >> behavior of another entry. I.e.: >> >> int y = 0; >> int x[2] = >> { >> y=1, /* first element will be always 1 */ >> y /* second element can be either 0 or 1 */ >> }; >> >> Above we have a dependency between the first entry and the second. >> >> Now let's consider logging effects: >> >> #define LOG(x) printf("LOG: %s", x); >> >> int x[2] = >> { >> ({ LOG("1"); 1; }), >> ({ LOG("2"); 2; }) >> }; >> >> >> Here the execution can print: >> "LOG: 1LOG: 2" or >> "LOG: 2LOG: 1". >> >> Do we agree that the evaluation order of prints caused by logging >> functions/macros do not normally cause dependencies between the >> entries? The execution is still non-deterministic, but does it really >> matter?. >> >> In the case of function that crash execution, no dependencies can exist >> since no further entries will be evaluated. >> >> In conclusion, I propose the following rewording: >> >> "text": "Effects that crash execution or produce logs are allowed in >> initializer lists. Logging effects do not affect the evaluation of >> subsequent entries. Crash effects are allowed as they represent the >> end of the execution. > > Let's assume we have a BUG_ON() (as the "crash effect") the condition of > which depends on where in the sequence of operations it actually executes, > i.e. (potentially) dependent upon another part of the initializer. I hope > we agree that's not something that should be deviated? Yet even the re- > worded statement would - according to my reading - permit doing so. > > I guess I should try to remember to bring this up on this afternoon's call. I known you had a conversation about that during the call, but from what I known, a clear decision for crash effects wasn't taken. If the community does not want SAF deviations for such cases, we should consider moving them to external variables. > >>>>>> --- a/xen/arch/arm/guestcopy.c >>>>>> +++ b/xen/arch/arm/guestcopy.c >>>>>> @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, >>>>>> unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) >>>>>> { >>>>>> return copy_guest((void *)from, (vaddr_t)to, len, >>>>>> + /* SAF-4-safe No persistent side effects */ >>>>>> GVA_INFO(current), COPY_to_guest | COPY_linear); >>>>>> } >>>>>> >>>>>> unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, >>>>>> unsigned int len) >>>>>> { >>>>>> + /* SAF-4-safe No persistent side effects */ >>>>>> return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), >>>>>> COPY_to_guest | COPY_flush_dcache | COPY_linear); >>>>>> } >>>>>> >>>>>> unsigned long raw_clear_guest(void *to, unsigned int len) >>>>>> { >>>>>> + /* SAF-4-safe No persistent side effects */ >>>>>> return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), >>>>>> COPY_to_guest | COPY_linear); >>>>>> } >>>>>> @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) >>>>>> unsigned long raw_copy_from_guest(void *to, const void __user *from, >>>>>> unsigned int len) >>>>>> { >>>>>> + /* SAF-4-safe No persistent side effects */ >>>>>> return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), >>>>>> COPY_from_guest | COPY_linear); >>>>>> } >>>>> >>>>> I can only guess that in all four of these it's the use of "current" which >>>>> requires the comment. Yet imo that either needs making explicit, or such a >>>>> comment shouldn't go on use sites of "current", but on its definition site. >>>>> >>>> >>>> "current" does not contain any violation of R13.1. Its expansion >>>> produces a side-effect, but this is not a problem in itself. The real >>>> problem is that GVA_INFO expands it in an initializer list: >>>> #define GVA_INFO(vcpu) ((copy_info_t) { .gva = { vcpu } }) >>> >>> But an initializer list doesn't itself constitute a side effect, does it? >> >> The side effect should be inside the initializer list. { .gva = 1 } is >> not a violation. > > I'm afraid I don't see what would be constituting a violation here. > > Jan >
diff --git a/docs/misra/safe.json b/docs/misra/safe.json index 952324f85c..69ab526084 100644 --- a/docs/misra/safe.json +++ b/docs/misra/safe.json @@ -28,6 +28,22 @@ }, { "id": "SAF-3-safe", + "analyser": { + "eclair": "MC3R1.R13.1" + }, + "name": "MC3R1.R13.1: effects for debugging and logging", + "text": "Effects for debugging and loggings reasons that crash execution or produce logs are allowed in initializer lists. The evaluation order in abnormal conditions is not relevant." + }, + { + "id": "SAF-4-safe", + "analyser": { + "eclair": "MC3R1.R13.1" + }, + "name": "MC3R1.R13.1: volatile asm statements that do not perform any persistent side effect", + "text": "Volatile asm statements in an initializer list if do not perform persistent side effects are safe." + }, + { + "id": "SAF-5-safe", "analyser": {}, "name": "Sentinel", "text": "Next ID to be used" diff --git a/xen/arch/arm/device.c b/xen/arch/arm/device.c index 1f631d3274..fa331f164d 100644 --- a/xen/arch/arm/device.c +++ b/xen/arch/arm/device.c @@ -331,6 +331,7 @@ int handle_device(struct domain *d, struct dt_device_node *dev, p2m_type_t p2mt, .p2mt = p2mt, .skip_mapping = !own_device || (is_pci_passthrough_enabled() && + /* SAF-3-safe effects for debugging/logging reasons are safe */ (device_get_class(dev) == DEVICE_PCI_HOSTBRIDGE)), .iomem_ranges = iomem_ranges, .irq_ranges = irq_ranges diff --git a/xen/arch/arm/guestcopy.c b/xen/arch/arm/guestcopy.c index 6716b03561..31b809ea08 100644 --- a/xen/arch/arm/guestcopy.c +++ b/xen/arch/arm/guestcopy.c @@ -110,18 +110,21 @@ static unsigned long copy_guest(void *buf, uint64_t addr, unsigned int len, unsigned long raw_copy_to_guest(void *to, const void *from, unsigned int len) { return copy_guest((void *)from, (vaddr_t)to, len, + /* SAF-4-safe No persistent side effects */ GVA_INFO(current), COPY_to_guest | COPY_linear); } unsigned long raw_copy_to_guest_flush_dcache(void *to, const void *from, unsigned int len) { + /* SAF-4-safe No persistent side effects */ return copy_guest((void *)from, (vaddr_t)to, len, GVA_INFO(current), COPY_to_guest | COPY_flush_dcache | COPY_linear); } unsigned long raw_clear_guest(void *to, unsigned int len) { + /* SAF-4-safe No persistent side effects */ return copy_guest(NULL, (vaddr_t)to, len, GVA_INFO(current), COPY_to_guest | COPY_linear); } @@ -129,6 +132,7 @@ unsigned long raw_clear_guest(void *to, unsigned int len) unsigned long raw_copy_from_guest(void *to, const void __user *from, unsigned int len) { + /* SAF-4-safe No persistent side effects */ return copy_guest(to, (vaddr_t)from, len, GVA_INFO(current), COPY_from_guest | COPY_linear); } diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c index 35a30df3b1..24fd02b363 100644 --- a/xen/arch/x86/hvm/hvm.c +++ b/xen/arch/x86/hvm/hvm.c @@ -1700,6 +1700,7 @@ void hvm_hlt(unsigned int eflags) do_sched_op(SCHEDOP_block, guest_handle_from_ptr(NULL, void)); + /* SAF-3-safe effects for debugging/logging reasons are safe */ HVMTRACE_1D(HLT, /* pending = */ vcpu_runnable(curr)); } diff --git a/xen/common/sched/core.c b/xen/common/sched/core.c index eba0cea4bb..b1b94c8689 100644 --- a/xen/common/sched/core.c +++ b/xen/common/sched/core.c @@ -1517,6 +1517,7 @@ long vcpu_yield(void) SCHED_STAT_CRANK(vcpu_yield); + /* SAF-4-safe No persistent side effects */ TRACE_2D(TRC_SCHED_YIELD, current->domain->domain_id, current->vcpu_id); raise_softirq(SCHEDULE_SOFTIRQ); return 0; @@ -1895,6 +1896,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) if ( copy_from_guest(&sched_shutdown, arg, 1) ) break; + /* SAF-4-safe No persistent side effects */ TRACE_3D(TRC_SCHED_SHUTDOWN, current->domain->domain_id, current->vcpu_id, sched_shutdown.reason); @@ -1912,6 +1914,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg) if ( copy_from_guest(&sched_shutdown, arg, 1) ) break; + /* SAF-4-safe No persistent side effects */ TRACE_3D(TRC_SCHED_SHUTDOWN_CODE, d->domain_id, current->vcpu_id, sched_shutdown.reason);
Rule 13.1: Initializer lists shall not contain persistent side effects Effects caused by debug/logging macros and functions (like ASSERT, __bad_atomic_size, LOG, etc ...) that crash execution or produce logs are not dangerous in initializer lists. The evaluation order in abnormal conditions is not relevant. Evaluation order of logging effects is always safe. This patch deviates violations using SAF commits caused by debug/logging macros and functions. Asm volatile statements in initializer lists that do not perform any persistent side effect are safe: this patch deviates violations caused by uses of the current macro (that contains an asm volatile) in initializer lists. No functional changes. Signed-off-by: Simone Ballarin <simone.ballarin@bugseng.com> --- Changes in v2: New patch based on the discussion for "xen/arm: address violations of MISRA C:2012 Rule 13.1". --- docs/misra/safe.json | 16 ++++++++++++++++ xen/arch/arm/device.c | 1 + xen/arch/arm/guestcopy.c | 4 ++++ xen/arch/x86/hvm/hvm.c | 1 + xen/common/sched/core.c | 3 +++ 5 files changed, 25 insertions(+)