lkml.org 
[lkml]   [2011]   [Nov]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    SubjectRe: [PATCH 1/1 v2] arch/arm/mm/fault.c: Porting OOM changes into __do_page_fault
    From
    Sorry, I made one more mistake about the locking:
    We cannot do a up_read as VM_FAULT_RETRY might be set.

    I will modify the patch and send it again for review.

    On Sat, Nov 12, 2011 at 7:51 PM, Kautuk Consul <consul.kautuk@gmail.com> wrote:
    > Commits d065bd810b6deb67d4897a14bfe21f8eb526ba99 and
    > 37b23e0525d393d48a7d59f870b3bc061a30ccdb introduced changes into
    > the x86 pagefault handler for makeing the page fault handler
    > retryable as well as killable.
    >
    > These changes reduce the mmap_sem hold time(for x86), which is crucial
    > during OOM killer invocation.
    >
    > Porting these changes to ARM.
    >
    > Without these changes, my ARM board encounters many hang and livelock
    > scenarios.
    > After applying this patch, OOM feature performance improves according to
    > my testing.
    >
    > Motivation for porting these changes:
    > ------------------------------------
    > Embedded devices such as SMART TVs and SMART phones in the near future
    > will have the capability to download and run apps from the internet.
    > Due to this, the device user might run some malignant app that
    > allocates too much memory.
    > In that case, OOM killer performance is very important so that the
    > device can free up memory for other apps as well as the kernel.
    >
    > Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
    > ---
    >  arch/arm/mm/fault.c |   48 +++++++++++++++++++++++++++++++++++++-----------
    >  1 files changed, 37 insertions(+), 11 deletions(-)
    >
    > diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c
    > index aa33949..7286786 100644
    > --- a/arch/arm/mm/fault.c
    > +++ b/arch/arm/mm/fault.c
    > @@ -231,7 +231,7 @@ static inline bool access_error(unsigned int fsr, struct vm_area_struct *vma)
    >
    >  static int __kprobes
    >  __do_page_fault(struct mm_struct *mm, unsigned long addr, unsigned int fsr,
    > -               struct task_struct *tsk)
    > +               unsigned int flags, struct task_struct *tsk)
    >  {
    >        struct vm_area_struct *vma;
    >        int fault;
    > @@ -257,13 +257,9 @@ good_area:
    >         * If for any reason at all we couldn't handle the fault, make
    >         * sure we exit gracefully rather than endlessly redo the fault.
    >         */
    > -       fault = handle_mm_fault(mm, vma, addr & PAGE_MASK, (fsr & FSR_WRITE) ? FAULT_FLAG_WRITE : 0);
    > +       fault = handle_mm_fault(mm, vma, addr & PAGE_MASK, flags);
    >        if (unlikely(fault & VM_FAULT_ERROR))
    >                return fault;
    > -       if (fault & VM_FAULT_MAJOR)
    > -               tsk->maj_flt++;
    > -       else
    > -               tsk->min_flt++;
    >        return fault;
    >
    >  check_stack:
    > @@ -279,6 +275,9 @@ do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
    >        struct task_struct *tsk;
    >        struct mm_struct *mm;
    >        int fault, sig, code;
    > +       int write = fsr & FSR_WRITE;
    > +       unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
    > +                                       (write ? FAULT_FLAG_WRITE : 0);
    >
    >        if (notify_page_fault(regs, fsr))
    >                return 0;
    > @@ -305,6 +304,7 @@ do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
    >        if (!down_read_trylock(&mm->mmap_sem)) {
    >                if (!user_mode(regs) && !search_exception_tables(regs->ARM_pc))
    >                        goto no_context;
    > +retry:
    >                down_read(&mm->mmap_sem);
    >        } else {
    >                /*
    > @@ -320,14 +320,40 @@ do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
    >  #endif
    >        }
    >
    > -       fault = __do_page_fault(mm, addr, fsr, tsk);
    > +       fault = __do_page_fault(mm, addr, fsr, flags, tsk);
    > +
    > +       /* If we need to retry but a fatal signal is pending, handle the
    > +        * signal first. We do not need to release the mmap_sem because
    > +        * it would already be released in __lock_page_or_retry in
    > +        * mm/filemap.c. */
    > +       if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
    > +               return 0;
    > +
    >        up_read(&mm->mmap_sem);
    >
    > +       /*
    > +        * Major/minor page fault accounting is only done on the
    > +        * initial attempt. If we go through a retry, it is extremely
    > +        * likely that the page will be found in page cache at that point.
    > +        */
    >        perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
    > -       if (fault & VM_FAULT_MAJOR)
    > -               perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, addr);
    > -       else if (fault & VM_FAULT_MINOR)
    > -               perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, addr);
    > +       if (flags & FAULT_FLAG_ALLOW_RETRY) {
    > +               if (fault & VM_FAULT_MAJOR) {
    > +                       tsk->maj_flt++;
    > +                       perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1,
    > +                                     regs, addr);
    > +               } else {
    > +                       tsk->min_flt++;
    > +                       perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1,
    > +                                     regs, addr);
    > +               }
    > +               if (fault & VM_FAULT_RETRY) {
    > +                       /* Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk
    > +                        * of starvation. */
    > +                       flags &= ~FAULT_FLAG_ALLOW_RETRY;
    > +                       goto retry;
    > +               }
    > +       }
    >
    >        /*
    >         * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR
    > --
    > 1.7.5.4
    >
    >
    --
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

    \
     
     \ /
      Last update: 2011-11-13 02:31    [W:0.041 / U:0.044 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site