| Message ID | 20200109134825.31482-2-jgross@suse.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | xen: fix CONFIG_DEBUG_LOCKS | expand |
On 09.01.2020 14:48, Juergen Gross wrote: > --- a/xen/Kconfig.debug > +++ b/xen/Kconfig.debug > @@ -81,6 +81,12 @@ config PERF_ARRAYS > ---help--- > Enables software performance counter array histograms. > > +config DEBUG_BUGVERBOSE > + bool "Verbose BUG_ON messages" > + default DEBUG > + ---help--- > + In case a BUG_ON triggers additionally print the triggering > + condition on the console. > > config VERBOSE_DEBUG While I can see reasons to put this here, doing so means the option will be unavailable in non-EXPERT release builds. Is it intended to be that way? > --- a/xen/include/xen/lib.h > +++ b/xen/include/xen/lib.h > @@ -8,7 +8,12 @@ > #include <xen/string.h> > #include <asm/bug.h> > > +#define BUG_ON_VERBOSE(p) do { if (unlikely(p)) BUG_VERBOSE(#p); } while (0) > +#ifdef CONFIG_DEBUG_BUGVERBOSE > +#define BUG_ON(p) BUG_ON_VERBOSE(p) Looks like this will fail to build on Arm? Also - any particular reason for the introduction of the separate BUG_ON_VERBOSE(), when BUG_ON() could directly use BUG_VERBOSE()? I don't think we want to encourage use of BUG_ON_VERBOSE() elsewhere ... Jan
On 14.01.20 16:47, Jan Beulich wrote: > On 09.01.2020 14:48, Juergen Gross wrote: >> --- a/xen/Kconfig.debug >> +++ b/xen/Kconfig.debug >> @@ -81,6 +81,12 @@ config PERF_ARRAYS >> ---help--- >> Enables software performance counter array histograms. >> >> +config DEBUG_BUGVERBOSE >> + bool "Verbose BUG_ON messages" >> + default DEBUG >> + ---help--- >> + In case a BUG_ON triggers additionally print the triggering >> + condition on the console. >> >> config VERBOSE_DEBUG > > While I can see reasons to put this here, doing so means the option > will be unavailable in non-EXPERT release builds. Is it intended to > be that way? I can move it either to xen/Kconfig or in Kconfig.debug out of the "if expert" section if you want. > >> --- a/xen/include/xen/lib.h >> +++ b/xen/include/xen/lib.h >> @@ -8,7 +8,12 @@ >> #include <xen/string.h> >> #include <asm/bug.h> >> >> +#define BUG_ON_VERBOSE(p) do { if (unlikely(p)) BUG_VERBOSE(#p); } while (0) >> +#ifdef CONFIG_DEBUG_BUGVERBOSE >> +#define BUG_ON(p) BUG_ON_VERBOSE(p) > > Looks like this will fail to build on Arm? Also - any particular Uh, shame on me! > reason for the introduction of the separate BUG_ON_VERBOSE(), > when BUG_ON() could directly use BUG_VERBOSE()? I don't think we > want to encourage use of BUG_ON_VERBOSE() elsewhere ... I wanted to offer that option. If you want me to remove it I wouldn't mind. Juergen
On 14.01.2020 17:00, Jürgen Groß wrote: > On 14.01.20 16:47, Jan Beulich wrote: >> On 09.01.2020 14:48, Juergen Gross wrote: >>> --- a/xen/Kconfig.debug >>> +++ b/xen/Kconfig.debug >>> @@ -81,6 +81,12 @@ config PERF_ARRAYS >>> ---help--- >>> Enables software performance counter array histograms. >>> >>> +config DEBUG_BUGVERBOSE >>> + bool "Verbose BUG_ON messages" >>> + default DEBUG >>> + ---help--- >>> + In case a BUG_ON triggers additionally print the triggering >>> + condition on the console. >>> >>> config VERBOSE_DEBUG >> >> While I can see reasons to put this here, doing so means the option >> will be unavailable in non-EXPERT release builds. Is it intended to >> be that way? > > I can move it either to xen/Kconfig or in Kconfig.debug out of the > "if expert" section if you want. I think this would be better, but give others a chance to voice opinions. >>> --- a/xen/include/xen/lib.h >>> +++ b/xen/include/xen/lib.h >>> @@ -8,7 +8,12 @@ >>> #include <xen/string.h> >>> #include <asm/bug.h> >>> >>> +#define BUG_ON_VERBOSE(p) do { if (unlikely(p)) BUG_VERBOSE(#p); } while (0) >>> +#ifdef CONFIG_DEBUG_BUGVERBOSE >>> +#define BUG_ON(p) BUG_ON_VERBOSE(p) >> >> Looks like this will fail to build on Arm? Also - any particular > > Uh, shame on me! > >> reason for the introduction of the separate BUG_ON_VERBOSE(), >> when BUG_ON() could directly use BUG_VERBOSE()? I don't think we >> want to encourage use of BUG_ON_VERBOSE() elsewhere ... > > I wanted to offer that option. If you want me to remove it I wouldn't > mind. As above - unless there are good reasons (making others to agree with you to have it), I'd prefer to not see it being independently usable, at least for the time being. Jan
On 14/01/2020 16:12, Jan Beulich wrote: > On 14.01.2020 17:00, Jürgen Groß wrote: >> On 14.01.20 16:47, Jan Beulich wrote: >>> On 09.01.2020 14:48, Juergen Gross wrote: >>>> --- a/xen/Kconfig.debug >>>> +++ b/xen/Kconfig.debug >>>> @@ -81,6 +81,12 @@ config PERF_ARRAYS >>>> ---help--- >>>> Enables software performance counter array histograms. >>>> >>>> +config DEBUG_BUGVERBOSE >>>> + bool "Verbose BUG_ON messages" >>>> + default DEBUG >>>> + ---help--- >>>> + In case a BUG_ON triggers additionally print the triggering >>>> + condition on the console. >>>> >>>> config VERBOSE_DEBUG >>> While I can see reasons to put this here, doing so means the option >>> will be unavailable in non-EXPERT release builds. Is it intended to >>> be that way? >> I can move it either to xen/Kconfig or in Kconfig.debug out of the >> "if expert" section if you want. > I think this would be better, but give others a chance to voice > opinions. TBH, I don't think anyone will be interested in not having the strings. The change is what? a couple of hundred bytes? That is a fraction of the size of some functions we have. Personally, I wouldn't even bother having the option. > >>>> --- a/xen/include/xen/lib.h >>>> +++ b/xen/include/xen/lib.h >>>> @@ -8,7 +8,12 @@ >>>> #include <xen/string.h> >>>> #include <asm/bug.h> >>>> >>>> +#define BUG_ON_VERBOSE(p) do { if (unlikely(p)) BUG_VERBOSE(#p); } while (0) >>>> +#ifdef CONFIG_DEBUG_BUGVERBOSE >>>> +#define BUG_ON(p) BUG_ON_VERBOSE(p) >>> Looks like this will fail to build on Arm? Also - any particular >> Uh, shame on me! >> >>> reason for the introduction of the separate BUG_ON_VERBOSE(), >>> when BUG_ON() could directly use BUG_VERBOSE()? I don't think we >>> want to encourage use of BUG_ON_VERBOSE() elsewhere ... >> I wanted to offer that option. If you want me to remove it I wouldn't >> mind. > As above - unless there are good reasons (making others to agree > with you to have it), I'd prefer to not see it being independently > usable, at least for the time being. I'd agree with the wish to not have a new flavour of BUG_ON(). People writing code aren't going to want the complexity of thinking about it, and people who care about the presence/absence of messages will care about it globally, not on a per-use bases. ~Andrew
On 16.01.2020 19:54, Andrew Cooper wrote: > On 14/01/2020 16:12, Jan Beulich wrote: >> On 14.01.2020 17:00, Jürgen Groß wrote: >>> On 14.01.20 16:47, Jan Beulich wrote: >>>> On 09.01.2020 14:48, Juergen Gross wrote: >>>>> --- a/xen/Kconfig.debug >>>>> +++ b/xen/Kconfig.debug >>>>> @@ -81,6 +81,12 @@ config PERF_ARRAYS >>>>> ---help--- >>>>> Enables software performance counter array histograms. >>>>> >>>>> +config DEBUG_BUGVERBOSE >>>>> + bool "Verbose BUG_ON messages" >>>>> + default DEBUG >>>>> + ---help--- >>>>> + In case a BUG_ON triggers additionally print the triggering >>>>> + condition on the console. >>>>> >>>>> config VERBOSE_DEBUG >>>> While I can see reasons to put this here, doing so means the option >>>> will be unavailable in non-EXPERT release builds. Is it intended to >>>> be that way? >>> I can move it either to xen/Kconfig or in Kconfig.debug out of the >>> "if expert" section if you want. >> I think this would be better, but give others a chance to voice >> opinions. > > TBH, I don't think anyone will be interested in not having the strings. > The change is what? a couple of hundred bytes? That is a fraction of > the size of some functions we have. Well, it's a couple thousand (about 7k according to my simplistic estimation, but this is taking Arm and x86 together). It's not the end of the world, but in particular embedded users may want to get rid of this. There's a reason after all (I assume) that Linux has the option. Jan
On 17.01.20 09:39, Jan Beulich wrote: > On 16.01.2020 19:54, Andrew Cooper wrote: >> On 14/01/2020 16:12, Jan Beulich wrote: >>> On 14.01.2020 17:00, Jürgen Groß wrote: >>>> On 14.01.20 16:47, Jan Beulich wrote: >>>>> On 09.01.2020 14:48, Juergen Gross wrote: >>>>>> --- a/xen/Kconfig.debug >>>>>> +++ b/xen/Kconfig.debug >>>>>> @@ -81,6 +81,12 @@ config PERF_ARRAYS >>>>>> ---help--- >>>>>> Enables software performance counter array histograms. >>>>>> >>>>>> +config DEBUG_BUGVERBOSE >>>>>> + bool "Verbose BUG_ON messages" >>>>>> + default DEBUG >>>>>> + ---help--- >>>>>> + In case a BUG_ON triggers additionally print the triggering >>>>>> + condition on the console. >>>>>> >>>>>> config VERBOSE_DEBUG >>>>> While I can see reasons to put this here, doing so means the option >>>>> will be unavailable in non-EXPERT release builds. Is it intended to >>>>> be that way? >>>> I can move it either to xen/Kconfig or in Kconfig.debug out of the >>>> "if expert" section if you want. >>> I think this would be better, but give others a chance to voice >>> opinions. >> >> TBH, I don't think anyone will be interested in not having the strings. >> The change is what? a couple of hundred bytes? That is a fraction of >> the size of some functions we have. > > Well, it's a couple thousand (about 7k according to my simplistic > estimation, but this is taking Arm and x86 together). It's not the > end of the world, but in particular embedded users may want to get > rid of this. There's a reason after all (I assume) that Linux has > the option. The needed code churn is rather limited, so I think we should keep the config option. Juergen
diff --git a/xen/Kconfig.debug b/xen/Kconfig.debug index b3511e81a2..dfbcac575a 100644 --- a/xen/Kconfig.debug +++ b/xen/Kconfig.debug @@ -81,6 +81,12 @@ config PERF_ARRAYS ---help--- Enables software performance counter array histograms. +config DEBUG_BUGVERBOSE + bool "Verbose BUG_ON messages" + default DEBUG + ---help--- + In case a BUG_ON triggers additionally print the triggering + condition on the console. config VERBOSE_DEBUG bool "Verbose debug messages" diff --git a/xen/include/asm-x86/bug.h b/xen/include/asm-x86/bug.h index 9bb4a19420..46d282777f 100644 --- a/xen/include/asm-x86/bug.h +++ b/xen/include/asm-x86/bug.h @@ -60,10 +60,11 @@ struct bug_frame { #define WARN() BUG_FRAME(BUGFRAME_warn, __LINE__, __FILE__, 0, NULL) -#define BUG() do { \ - BUG_FRAME(BUGFRAME_bug, __LINE__, __FILE__, 0, NULL); \ +#define BUG_VERBOSE(msg) do { \ + BUG_FRAME(BUGFRAME_bug, __LINE__, __FILE__, 0, msg); \ unreachable(); \ } while (0) +#define BUG() BUG_VERBOSE(NULL) #define run_in_exception_handler(fn) BUG_FRAME(BUGFRAME_run_fn, 0, fn, 0, NULL) diff --git a/xen/include/xen/lib.h b/xen/include/xen/lib.h index 8fbe84032d..e7770b0d24 100644 --- a/xen/include/xen/lib.h +++ b/xen/include/xen/lib.h @@ -8,7 +8,12 @@ #include <xen/string.h> #include <asm/bug.h> +#define BUG_ON_VERBOSE(p) do { if (unlikely(p)) BUG_VERBOSE(#p); } while (0) +#ifdef CONFIG_DEBUG_BUGVERBOSE +#define BUG_ON(p) BUG_ON_VERBOSE(p) +#else #define BUG_ON(p) do { if (unlikely(p)) BUG(); } while (0) +#endif #define WARN_ON(p) do { if (unlikely(p)) WARN(); } while (0) #if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)
Today a triggering BUG_ON() will only print source file and line information. Add the possibility to print the triggering condition like ASSERT(). Do that by introducing BUG_ON_VERBOSE() and add a Kconfig option to make BUG_ON use BUG_ON_VERBOSE(). Signed-off-by: Juergen Gross <jgross@suse.com> --- xen/Kconfig.debug | 6 ++++++ xen/include/asm-x86/bug.h | 5 +++-- xen/include/xen/lib.h | 5 +++++ 3 files changed, 14 insertions(+), 2 deletions(-)