mm: trim __do_fault() arguments
authorJan Kara <jack@suse.cz>
Wed, 14 Dec 2016 23:07:10 +0000 (15:07 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 15 Dec 2016 00:04:09 +0000 (16:04 -0800)
Use vm_fault structure to pass cow_page, page, and entry in and out of
the function.

That reduces number of __do_fault() arguments from 4 to 1.

Link: http://lkml.kernel.org/r/1479460644-25076-6-git-send-email-jack@suse.cz
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memory.c

index cbc6d47fda7350e32017639ebb6dfc169bd85fba..78b81e8984df677df5ed3ba1cbb8ddc83b4b4111 100644 (file)
@@ -2844,26 +2844,22 @@ oom:
  * released depending on flags and vma->vm_ops->fault() return value.
  * See filemap_fault() and __lock_page_retry().
  */
-static int __do_fault(struct vm_fault *vmf, struct page *cow_page,
-                     struct page **page, void **entry)
+static int __do_fault(struct vm_fault *vmf)
 {
        struct vm_area_struct *vma = vmf->vma;
        int ret;
 
-       vmf->cow_page = cow_page;
-
        ret = vma->vm_ops->fault(vma, vmf);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
                return ret;
-       if (ret & VM_FAULT_DAX_LOCKED) {
-               *entry = vmf->entry;
+       if (ret & VM_FAULT_DAX_LOCKED)
                return ret;
-       }
 
        if (unlikely(PageHWPoison(vmf->page))) {
                if (ret & VM_FAULT_LOCKED)
                        unlock_page(vmf->page);
                put_page(vmf->page);
+               vmf->page = NULL;
                return VM_FAULT_HWPOISON;
        }
 
@@ -2872,7 +2868,6 @@ static int __do_fault(struct vm_fault *vmf, struct page *cow_page,
        else
                VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
 
-       *page = vmf->page;
        return ret;
 }
 
@@ -3208,7 +3203,6 @@ out:
 static int do_read_fault(struct vm_fault *vmf)
 {
        struct vm_area_struct *vma = vmf->vma;
-       struct page *fault_page;
        int ret = 0;
 
        /*
@@ -3222,54 +3216,52 @@ static int do_read_fault(struct vm_fault *vmf)
                        return ret;
        }
 
-       ret = __do_fault(vmf, NULL, &fault_page, NULL);
+       ret = __do_fault(vmf);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
                return ret;
 
-       ret |= alloc_set_pte(vmf, NULL, fault_page);
+       ret |= alloc_set_pte(vmf, NULL, vmf->page);
        if (vmf->pte)
                pte_unmap_unlock(vmf->pte, vmf->ptl);
-       unlock_page(fault_page);
+       unlock_page(vmf->page);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
-               put_page(fault_page);
+               put_page(vmf->page);
        return ret;
 }
 
 static int do_cow_fault(struct vm_fault *vmf)
 {
        struct vm_area_struct *vma = vmf->vma;
-       struct page *fault_page, *new_page;
-       void *fault_entry;
        struct mem_cgroup *memcg;
        int ret;
 
        if (unlikely(anon_vma_prepare(vma)))
                return VM_FAULT_OOM;
 
-       new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
-       if (!new_page)
+       vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
+       if (!vmf->cow_page)
                return VM_FAULT_OOM;
 
-       if (mem_cgroup_try_charge(new_page, vma->vm_mm, GFP_KERNEL,
+       if (mem_cgroup_try_charge(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
                                &memcg, false)) {
-               put_page(new_page);
+               put_page(vmf->cow_page);
                return VM_FAULT_OOM;
        }
 
-       ret = __do_fault(vmf, new_page, &fault_page, &fault_entry);
+       ret = __do_fault(vmf);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
                goto uncharge_out;
 
        if (!(ret & VM_FAULT_DAX_LOCKED))
-               copy_user_highpage(new_page, fault_page, vmf->address, vma);
-       __SetPageUptodate(new_page);
+               copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
+       __SetPageUptodate(vmf->cow_page);
 
-       ret |= alloc_set_pte(vmf, memcg, new_page);
+       ret |= alloc_set_pte(vmf, memcg, vmf->cow_page);
        if (vmf->pte)
                pte_unmap_unlock(vmf->pte, vmf->ptl);
        if (!(ret & VM_FAULT_DAX_LOCKED)) {
-               unlock_page(fault_page);
-               put_page(fault_page);
+               unlock_page(vmf->page);
+               put_page(vmf->page);
        } else {
                dax_unlock_mapping_entry(vma->vm_file->f_mapping, vmf->pgoff);
        }
@@ -3277,20 +3269,19 @@ static int do_cow_fault(struct vm_fault *vmf)
                goto uncharge_out;
        return ret;
 uncharge_out:
-       mem_cgroup_cancel_charge(new_page, memcg, false);
-       put_page(new_page);
+       mem_cgroup_cancel_charge(vmf->cow_page, memcg, false);
+       put_page(vmf->cow_page);
        return ret;
 }
 
 static int do_shared_fault(struct vm_fault *vmf)
 {
        struct vm_area_struct *vma = vmf->vma;
-       struct page *fault_page;
        struct address_space *mapping;
        int dirtied = 0;
        int ret, tmp;
 
-       ret = __do_fault(vmf, NULL, &fault_page, NULL);
+       ret = __do_fault(vmf);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
                return ret;
 
@@ -3299,26 +3290,26 @@ static int do_shared_fault(struct vm_fault *vmf)
         * about to become writable
         */
        if (vma->vm_ops->page_mkwrite) {
-               unlock_page(fault_page);
-               tmp = do_page_mkwrite(vma, fault_page, vmf->address);
+               unlock_page(vmf->page);
+               tmp = do_page_mkwrite(vma, vmf->page, vmf->address);
                if (unlikely(!tmp ||
                                (tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
-                       put_page(fault_page);
+                       put_page(vmf->page);
                        return tmp;
                }
        }
 
-       ret |= alloc_set_pte(vmf, NULL, fault_page);
+       ret |= alloc_set_pte(vmf, NULL, vmf->page);
        if (vmf->pte)
                pte_unmap_unlock(vmf->pte, vmf->ptl);
        if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
                                        VM_FAULT_RETRY))) {
-               unlock_page(fault_page);
-               put_page(fault_page);
+               unlock_page(vmf->page);
+               put_page(vmf->page);
                return ret;
        }
 
-       if (set_page_dirty(fault_page))
+       if (set_page_dirty(vmf->page))
                dirtied = 1;
        /*
         * Take a local copy of the address_space - page.mapping may be zeroed
@@ -3326,8 +3317,8 @@ static int do_shared_fault(struct vm_fault *vmf)
         * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
         * release semantics to prevent the compiler from undoing this copying.
         */
-       mapping = page_rmapping(fault_page);
-       unlock_page(fault_page);
+       mapping = page_rmapping(vmf->page);
+       unlock_page(vmf->page);
        if ((dirtied || vma->vm_ops->page_mkwrite) && mapping) {
                /*
                 * Some device drivers do not set page.mapping but still