Re: [PATCH v6 2/2] arm64/crash_core: Export TCR_EL1.T1SZ in vmcoreinfo

[Will Deacon <will@kernel.org>]

On Thu, Jun 04, 2020 at 02:04:58AM +0530, Bhupesh Sharma wrote:
> On Wed, Jun 3, 2020 at 4:50 PM Kamlakant Patel <kamlakantp@marvell.com> wrote:
> > > diff --git a/arch/arm64/kernel/crash_core.c b/arch/arm64/kernel/crash_core.c
> > > index 1f646b07e3e9..314391a156ee 100644
> > > --- a/arch/arm64/kernel/crash_core.c
> > > +++ b/arch/arm64/kernel/crash_core.c
> > > @@ -7,6 +7,14 @@
> > >  #include <linux/crash_core.h>
> > >  #include <asm/cpufeature.h>
> > >  #include <asm/memory.h>
> > > +#include <asm/pgtable-hwdef.h>
> > > +
> > > +static inline u64 get_tcr_el1_t1sz(void);
> > > +
> > > +static inline u64 get_tcr_el1_t1sz(void) {
> > > +     return (read_sysreg(tcr_el1) & TCR_T1SZ_MASK) >> TCR_T1SZ_OFFSET; }
> > >
> > >  void arch_crash_save_vmcoreinfo(void)
> > >  {
> > > @@ -16,6 +24,8 @@ void arch_crash_save_vmcoreinfo(void)
> > >                                               kimage_voffset);
> > >       vmcoreinfo_append_str("NUMBER(PHYS_OFFSET)=0x%llx\n",
> > >                                               PHYS_OFFSET);
> > > +     vmcoreinfo_append_str("NUMBER(TCR_EL1_T1SZ)=0x%llx\n",
> > > +                                             get_tcr_el1_t1sz());
> > I tested this patch on top of upstream kernel v5.7 and I am getting "crash: cannot determine VA_BITS_ACTUAL" error with crash tool.
> > I looked into crash-utility source and it is expecting tcr_el1_t1sz not TCR_EL1_T1SZ.
> > Could you please check.
> 
> Indeed. As per James comments on the v5 (see [1]) where he suggested
> converting ttcr_el1_t1sz into TCR_EL1_T1SZ, I made the change in v6
> accordingly.
> 
> This time I haven't sent out the v6 userspace changes
> (makedumpfile/crash-utility) upstream first, since we are waiting for
> kernel changes to be accepted first, as we have seen in the past that
> while the userspace patches have been accepted, the kernel patches
> required a respin cycle, thus leading to inconsistencies, as you also
> pointed out with crash-utility.

Yes, and that';s the right way to do it. Userspace can't rely on the
stability of a kernel interface if it's no in the upstream kernel!

So doing with the ALL CAPS names is the right thing to do.

Will