Re: [PATCH V2] arm64/cpufeature: Add get_arm64_ftr_reg_nowarn()

[Will Deacon]

On Wed, May 27, 2020 at 07:56:30AM +0530, Anshuman Khandual wrote:
> 
> 
> On 05/27/2020 01:16 AM, Will Deacon wrote:
> > On Tue, May 26, 2020 at 04:01:35PM +0100, Catalin Marinas wrote:
> >> On Tue, May 26, 2020 at 07:09:13PM +0530, Anshuman Khandual wrote:
> >>> @@ -632,8 +654,6 @@ static void __init init_cpu_ftr_reg(u32 sys_reg, u64 new)
> >>>  	const struct arm64_ftr_bits *ftrp;
> >>>  	struct arm64_ftr_reg *reg = get_arm64_ftr_reg(sys_reg);
> >>>  
> >>> -	BUG_ON(!reg);
> >>> -
> >>>  	for (ftrp = reg->ftr_bits; ftrp->width; ftrp++) {
> >>>  		u64 ftr_mask = arm64_ftr_mask(ftrp);
> >>>  		s64 ftr_new = arm64_ftr_value(ftrp, new);
> >>> @@ -762,7 +782,6 @@ static int check_update_ftr_reg(u32 sys_id, int cpu, u64 val, u64 boot)
> >>>  {
> >>>  	struct arm64_ftr_reg *regp = get_arm64_ftr_reg(sys_id);
> >>>  
> >>> -	BUG_ON(!regp);
> >>>  	update_cpu_ftr_reg(regp, val);
> >>>  	if ((boot & regp->strict_mask) == (val & regp->strict_mask))
> >>>  		return 0;
> >>> @@ -776,9 +795,6 @@ static void relax_cpu_ftr_reg(u32 sys_id, int field)
> >>>  	const struct arm64_ftr_bits *ftrp;
> >>>  	struct arm64_ftr_reg *regp = get_arm64_ftr_reg(sys_id);
> >>>  
> >>> -	if (WARN_ON(!regp))
> >>> -		return;
> >>
> >> I think Will wanted an early return in all these functions not just
> >> removing the BUG_ON(). I'll let him clarify.
> > 
> > Yes, the callers need to check the pointer and return early.
> 
> Sure, will do. But for check_update_ftr_reg(), a feature register search
> failure should be treated as a success (0) or a failure (1). What should
> it return ? Seems bit tricky, as there are good reasons to go either way.

We're unable to check it so return 0, otherwise we'll randomly taint the
kernel and print a weird message.

Will