Re: [PATCH v6 6/7] kgdb: roundup: Allow runtime arch specific override

[Daniel Thompson <daniel.thompson@linaro.org>]

On Mon, Nov 02, 2020 at 11:48:53AM +0530, Sumit Garg wrote:
> On Thu, 29 Oct 2020 at 20:51, Daniel Thompson
> <daniel.thompson@linaro.org> wrote:
> >
> > On Thu, Oct 29, 2020 at 08:26:26PM +0530, Sumit Garg wrote:
> > > Add a new API kgdb_arch_roundup_cpus() for a particular archichecture to
> > > override default kgdb roundup and if it detects at runtime to not support
> > > NMI roundup then it can fallback to default implementation using async
> > > SMP cross-calls.
> > >
> > > Currently such an architecture example is arm64 supporting pseudo NMIs
> > > feature which is only available on platforms which have support for GICv3
> > > or later version.
> > >
> > > Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> > > ---
> > >  arch/powerpc/kernel/kgdb.c |  3 ++-
> > >  arch/sparc/kernel/smp_64.c |  3 ++-
> > >  arch/x86/kernel/kgdb.c     |  6 ++++--
> > >  include/linux/kgdb.h       |  5 +++--
> > >  kernel/debug/debug_core.c  | 10 +++++++++-
> > >  5 files changed, 20 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/arch/powerpc/kernel/kgdb.c b/arch/powerpc/kernel/kgdb.c
> > > index 4090802..126575d 100644
> > > --- a/arch/powerpc/kernel/kgdb.c
> > > +++ b/arch/powerpc/kernel/kgdb.c
> > > @@ -125,9 +125,10 @@ static int kgdb_debugger_ipi(struct pt_regs *regs)
> > >  }
> > >
> > >  #ifdef CONFIG_SMP
> > > -void kgdb_roundup_cpus(void)
> > > +bool kgdb_arch_roundup_cpus(void)
> > >  {
> > >       smp_send_debugger_break();
> > > +     return true;
> > >  }
> > >  #endif
> > >
> > > diff --git a/arch/sparc/kernel/smp_64.c b/arch/sparc/kernel/smp_64.c
> > > index e38d8bf..c459c83 100644
> > > --- a/arch/sparc/kernel/smp_64.c
> > > +++ b/arch/sparc/kernel/smp_64.c
> > > @@ -1014,9 +1014,10 @@ void flush_dcache_page_all(struct mm_struct *mm, struct page *page)
> > >  }
> > >
> > >  #ifdef CONFIG_KGDB
> > > -void kgdb_roundup_cpus(void)
> > > +bool kgdb_arch_roundup_cpus(void)
> > >  {
> > >       smp_cross_call(&xcall_kgdb_capture, 0, 0, 0);
> > > +     return true;
> > >  }
> > >  #endif
> > >
> > > diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
> > > index ff7878d..1b756d9 100644
> > > --- a/arch/x86/kernel/kgdb.c
> > > +++ b/arch/x86/kernel/kgdb.c
> > > @@ -404,7 +404,8 @@ static void kgdb_disable_hw_debug(struct pt_regs *regs)
> > >
> > >  #ifdef CONFIG_SMP
> > >  /**
> > > - *   kgdb_roundup_cpus - Get other CPUs into a holding pattern
> > > + *   kgdb_arch_roundup_cpus - Get other CPUs into a holding pattern
> > > + *                            in an architectural specific manner
> > >   *
> > >   *   On SMP systems, we need to get the attention of the other CPUs
> > >   *   and get them be in a known state.  This should do what is needed
> > > @@ -414,9 +415,10 @@ static void kgdb_disable_hw_debug(struct pt_regs *regs)
> > >   *
> > >   *   On non-SMP systems, this is not called.
> > >   */
> > > -void kgdb_roundup_cpus(void)
> > > +bool kgdb_arch_roundup_cpus(void)
> > >  {
> > >       apic_send_IPI_allbutself(NMI_VECTOR);
> > > +     return true;
> > >  }
> > >  #endif
> > >
> > > diff --git a/include/linux/kgdb.h b/include/linux/kgdb.h
> > > index 0d6cf64..f9db5b8 100644
> > > --- a/include/linux/kgdb.h
> > > +++ b/include/linux/kgdb.h
> > > @@ -200,7 +200,8 @@ kgdb_arch_handle_qxfer_pkt(char *remcom_in_buffer,
> > >  extern void kgdb_call_nmi_hook(void *ignored);
> > >
> > >  /**
> > > - *   kgdb_roundup_cpus - Get other CPUs into a holding pattern
> > > + *   kgdb_arch_roundup_cpus - Get other CPUs into a holding pattern
> > > + *                            in an architectural specific manner
> > >   *
> > >   *   On SMP systems, we need to get the attention of the other CPUs
> > >   *   and get them into a known state.  This should do what is needed
> > > @@ -210,7 +211,7 @@ extern void kgdb_call_nmi_hook(void *ignored);
> > >   *
> > >   *   On non-SMP systems, this is not called.
> > >   */
> > > -extern void kgdb_roundup_cpus(void);
> > > +extern bool kgdb_arch_roundup_cpus(void);
> > >
> > >  /**
> > >   *   kgdb_arch_set_pc - Generic call back to the program counter
> > > diff --git a/kernel/debug/debug_core.c b/kernel/debug/debug_core.c
> > > index 1e75a89..27e401c 100644
> > > --- a/kernel/debug/debug_core.c
> > > +++ b/kernel/debug/debug_core.c
> > > @@ -241,13 +241,21 @@ void __weak kgdb_call_nmi_hook(void *ignored)
> > >  }
> > >  NOKPROBE_SYMBOL(kgdb_call_nmi_hook);
> > >
> > > -void __weak kgdb_roundup_cpus(void)
> > > +bool __weak kgdb_arch_roundup_cpus(void)
> > > +{
> > > +     return false;
> >
> > Do we really need to introduce all these boolean return values just to
> > call a bit of library code when one of the architectures wants to
> > fallback to a generic implementation?
> >
> > Why not something more like:
> >
> > void kgdb_smp_call_nmi_hook(void)
> > {
> >     /* original weak version of kgdb_roundup_cpus goes here */
> > }
> >
> > void __weak kgdb_roundup_cpus(void)
> > {
> >     kgdb_smp_call_nmi_hook();
> > }
> >
> > arm64 can now simply call the new library function if a fallback is needed.
> >
> 
> Makes sense, I will use this approach instead.
> 
> > Note that same question applies to the backtrace changes...
> 
> In case of backtrace, there are already multiple APIs wrapping
> arch_trigger_cpumask_backtrace() as follows:
> 
> - trigger_all_cpu_backtrace()
> - trigger_allbutself_cpu_backtrace()
> - trigger_cpumask_backtrace()
> - trigger_single_cpu_backtrace()
> 
> And each of them already return a boolean and have corresponding
> different fallback mechanisms. So we can't have a common fallback API
> from arch specific code and instead need to extend that boolean return
> to arch specific code that is implemented as part of patch #4.

Understood. Thanks.


Daniel.