Re: [PATCH v2 2/3] arm64: hwpoison: add VM_FAULT_HWPOISON[_LARGE] handling

[Will Deacon]

On Wed, May 17, 2017 at 04:23:35PM +0100, Punit Agrawal wrote:
> From: "Jonathan (Zhixiong) Zhang" <zjzhang@xxxxxxxxxxxxxx>
> 
> Add VM_FAULT_HWPOISON[_LARGE] handling to the arm64 page fault
> handler. Handling of VM_FAULT_HWPOISON[_LARGE] is very similar
> to VM_FAULT_OOM, the only difference is that a different si_code
> (BUS_MCEERR_AR) is passed to user space and si_addr_lsb field is
> initialized.
> 
> Signed-off-by: Jonathan (Zhixiong) Zhang <zjzhang@xxxxxxxxxxxxxx>
> Signed-off-by: Tyler Baicar <tbaicar@xxxxxxxxxxxxxx>
> (fix new __do_user_fault call-site)
> Signed-off-by: Punit Agrawal <punit.agrawal@xxxxxxx>
> Acked-by: Steve Capper <steve.capper@xxxxxxx>
> ---
>  arch/arm64/mm/fault.c | 22 +++++++++++++++++++---
>  1 file changed, 19 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
> index 37b95dff0b07..a85b44343ac6 100644
> --- a/arch/arm64/mm/fault.c
> +++ b/arch/arm64/mm/fault.c
> @@ -31,6 +31,7 @@
>  #include <linux/highmem.h>
>  #include <linux/perf_event.h>
>  #include <linux/preempt.h>
> +#include <linux/hugetlb.h>
>  
>  #include <asm/bug.h>
>  #include <asm/cpufeature.h>
> @@ -239,10 +240,11 @@ static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
>   */
>  static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
>  			    unsigned int esr, unsigned int sig, int code,
> -			    struct pt_regs *regs)
> +			    struct pt_regs *regs, int fault)
>  {
>  	struct siginfo si;
>  	const struct fault_info *inf;
> +	unsigned int lsb = 0;
>  
>  	if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
>  		inf = esr_to_fault_info(esr);
> @@ -259,6 +261,17 @@ static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
>  	si.si_errno = 0;
>  	si.si_code = code;
>  	si.si_addr = (void __user *)addr;
> +	/*
> +	 * Either small page or large page may be poisoned.
> +	 * In other words, VM_FAULT_HWPOISON_LARGE and
> +	 * VM_FAULT_HWPOISON are mutually exclusive.
> +	 */
> +	if (fault & VM_FAULT_HWPOISON_LARGE)
> +		lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
> +	else if (fault & VM_FAULT_HWPOISON)
> +		lsb = PAGE_SHIFT;
> +	si.si_addr_lsb = lsb;
> +

If we're going to start handling poison faults, then we should probably
rejig the perf page fault accounting around here so that we follow x86:

  * Always report PERF_COUNT_SW_PAGE_FAULTS,
  * Don't report anything else for VM_FAULT_ERROR
  * Report PERF_COUNT_SW_PAGE_FAULTS_MAJ if VM_FAULT_MAJOR
  * Otherwise, report PERF_COUNT_SW_PAGE_FAULTS_MIN

at the moment, I think you're accounting VM_FAULT_ERROR as
PERF_COUNT_SW_PAGE_FAULTS_MIN, which doesn't feel right at all.

Will