Introduce vma_can_speculate(), which allows speculative handling for
VMAs mapping supported file types.

>From do_handle_mm_fault(), speculative handling will follow through
__handle_mm_fault(), handle_pte_fault() and do_fault().

At this point, we expect speculative faults to continue through one of:
- do_read_fault(), fully implemented;
- do_cow_fault(), which might abort if missing anon vmas,
- do_shared_fault(), not implemented yet
  (would require ->page_mkwrite() changes).

vma_can_speculate() provides an early abort for the do_shared_fault() case,
limiting the time spent on trying that unimplemented case.

Signed-off-by: Michel Lespinasse <mic...@lespinasse.org>
---
 arch/x86/mm/fault.c |  3 ++-
 include/linux/mm.h  | 14 ++++++++++++++
 mm/memory.c         | 17 ++++++++++++-----
 3 files changed, 28 insertions(+), 6 deletions(-)

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index fbf265f56a06..48b86911a6df 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -1324,7 +1324,8 @@ void do_user_addr_fault(struct pt_regs *regs,
                goto spf_abort;
        rcu_read_lock();
        vma = find_vma(mm, address);
-       if (!vma || vma->vm_start > address || !vma_is_anonymous(vma)) {
+       if (!vma || vma->vm_start > address ||
+           !vma_can_speculate(vma, flags)) {
                rcu_read_unlock();
                goto spf_abort;
        }
diff --git a/include/linux/mm.h b/include/linux/mm.h
index b4c0c10e434e..edb809e9036b 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -700,6 +700,20 @@ static inline bool vma_is_accessible(struct vm_area_struct 
*vma)
        return vma->vm_flags & VM_ACCESS_FLAGS;
 }
 
+static inline bool vma_can_speculate(struct vm_area_struct *vma,
+               unsigned int flags)
+{
+       if (vma_is_anonymous(vma))
+               return true;
+       if (!vma->vm_ops->speculative)
+               return false;
+       if (!(flags & FAULT_FLAG_WRITE))
+               return true;
+       if (!(vma->vm_flags & VM_SHARED))
+               return true;
+       return false;
+}
+
 #ifdef CONFIG_SHMEM
 /*
  * The vma_is_shmem is not inline because it is used only by slow
diff --git a/mm/memory.c b/mm/memory.c
index a20e13d84145..074945faf1ab 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -4090,8 +4090,12 @@ static vm_fault_t do_cow_fault(struct vm_fault *vmf)
        struct vm_area_struct *vma = vmf->vma;
        vm_fault_t ret;
 
-       if (unlikely(anon_vma_prepare(vma)))
-               return VM_FAULT_OOM;
+       if (unlikely(!vma->anon_vma)) {
+               if (vmf->flags & FAULT_FLAG_SPECULATIVE)
+                       return VM_FAULT_RETRY;
+               if (__anon_vma_prepare(vma))
+                       return VM_FAULT_OOM;
+       }
 
        vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
        if (!vmf->cow_page)
@@ -4128,6 +4132,8 @@ static vm_fault_t do_shared_fault(struct vm_fault *vmf)
        struct vm_area_struct *vma = vmf->vma;
        vm_fault_t ret, tmp;
 
+       VM_BUG_ON(vmf->flags & FAULT_FLAG_SPECULATIVE);
+
        ret = __do_fault(vmf);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
                return ret;
@@ -4172,12 +4178,12 @@ static vm_fault_t do_fault(struct vm_fault *vmf)
        struct mm_struct *vm_mm = vma->vm_mm;
        vm_fault_t ret;
 
-       VM_BUG_ON(vmf->flags & FAULT_FLAG_SPECULATIVE);
-
        /*
         * The VMA was not fully populated on mmap() or missing VM_DONTEXPAND
         */
        if (!vma->vm_ops->fault) {
+               VM_BUG_ON(vmf->flags & FAULT_FLAG_SPECULATIVE);
+
                /*
                 * If we find a migration pmd entry or a none pmd entry, which
                 * should never happen, return SIGBUS
@@ -4739,7 +4745,8 @@ vm_fault_t do_handle_mm_fault(struct vm_area_struct *vma,
 {
        vm_fault_t ret;
 
-       VM_BUG_ON((flags & FAULT_FLAG_SPECULATIVE) && !vma_is_anonymous(vma));
+       VM_BUG_ON((flags & FAULT_FLAG_SPECULATIVE) &&
+                 !vma_can_speculate(vma, flags));
 
        __set_current_state(TASK_RUNNING);
 
-- 
2.20.1

Reply via email to