diff mbox series

[v6,4/6] mm: change folio_lock_or_retry to use vm_fault directly

Message ID 20230630020436.1066016-5-surenb@google.com (mailing list archive)
State New, archived
Headers show
Series Per-VMA lock support for swap and userfaults | expand

Commit Message

Suren Baghdasaryan June 30, 2023, 2:04 a.m. UTC 
Change folio_lock_or_retry to accept vm_fault struct and return the
vm_fault_t directly.

Suggested-by: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
---
 include/linux/pagemap.h |  9 ++++-----
 mm/filemap.c            | 22 ++++++++++++----------
 mm/memory.c             | 14 ++++++--------
 3 files changed, 22 insertions(+), 23 deletions(-)

Comments

Matthew Wilcox (Oracle) June 30, 2023, 3:36 a.m. UTC | #1 
On Thu, Jun 29, 2023 at 07:04:33PM -0700, Suren Baghdasaryan wrote:
> Change folio_lock_or_retry to accept vm_fault struct and return the
> vm_fault_t directly.

I thought we decided to call this folio_lock_fault()?

> +static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
> +					     struct vm_fault *vmf)
>  {
>  	might_sleep();
> -	return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
> +	return folio_trylock(folio) ? 0 : __folio_lock_or_retry(folio, vmf);

No, don't use the awful ternary operator.  The || form is used
everywhere else.

>  /*
>   * Return values:
> - * true - folio is locked; mmap_lock is still held.
> - * false - folio is not locked.
> + * 0 - folio is locked.
> + * VM_FAULT_RETRY - folio is not locked.

I don't think we want to be so prescriptive here.  It returns non-zero
if the folio is not locked.  The precise value is not something that
callers should depend on.
Suren Baghdasaryan June 30, 2023, 3:45 a.m. UTC | #2 
On Thu, Jun 29, 2023 at 8:36 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Thu, Jun 29, 2023 at 07:04:33PM -0700, Suren Baghdasaryan wrote:
> > Change folio_lock_or_retry to accept vm_fault struct and return the
> > vm_fault_t directly.
>
> I thought we decided to call this folio_lock_fault()?
>
> > +static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
> > +                                          struct vm_fault *vmf)
> >  {
> >       might_sleep();
> > -     return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
> > +     return folio_trylock(folio) ? 0 : __folio_lock_or_retry(folio, vmf);
>
> No, don't use the awful ternary operator.  The || form is used
> everywhere else.

Ok, but folio_trylock() returns a boolean while folio_lock_or_retry
should return vm_fault_t. How exactly do you suggest changing this?
Something like this perhaps:

static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
                                          struct vm_fault *vmf)
{
     might_sleep();
     if (folio_trylock(folio))
         return 0;
     return __folio_lock_or_retry(folio, mm, flags);
}

?


>
> >  /*
> >   * Return values:
> > - * true - folio is locked; mmap_lock is still held.
> > - * false - folio is not locked.
> > + * 0 - folio is locked.
> > + * VM_FAULT_RETRY - folio is not locked.
>
> I don't think we want to be so prescriptive here.  It returns non-zero
> if the folio is not locked.  The precise value is not something that
> callers should depend on.

Ok, I'll change it to "non-zero" here.

>
Matthew Wilcox (Oracle) June 30, 2023, 3:47 a.m. UTC | #3 
On Thu, Jun 29, 2023 at 08:45:39PM -0700, Suren Baghdasaryan wrote:
> On Thu, Jun 29, 2023 at 8:36 PM Matthew Wilcox <willy@infradead.org> wrote:
> >
> > On Thu, Jun 29, 2023 at 07:04:33PM -0700, Suren Baghdasaryan wrote:
> > > Change folio_lock_or_retry to accept vm_fault struct and return the
> > > vm_fault_t directly.
> >
> > I thought we decided to call this folio_lock_fault()?
> >
> > > +static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
> > > +                                          struct vm_fault *vmf)
> > >  {
> > >       might_sleep();
> > > -     return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
> > > +     return folio_trylock(folio) ? 0 : __folio_lock_or_retry(folio, vmf);
> >
> > No, don't use the awful ternary operator.  The || form is used
> > everywhere else.
> 
> Ok, but folio_trylock() returns a boolean while folio_lock_or_retry
> should return vm_fault_t. How exactly do you suggest changing this?
> Something like this perhaps:
> 
> static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
>                                           struct vm_fault *vmf)
> {
>      might_sleep();
>      if (folio_trylock(folio))
>          return 0;
>      return __folio_lock_or_retry(folio, mm, flags);
> }
> 
> ?

I think the automatic casting would work, but I prefer what you've
written here.

> > >  /*
> > >   * Return values:
> > > - * true - folio is locked; mmap_lock is still held.
> > > - * false - folio is not locked.
> > > + * 0 - folio is locked.
> > > + * VM_FAULT_RETRY - folio is not locked.
> >
> > I don't think we want to be so prescriptive here.  It returns non-zero
> > if the folio is not locked.  The precise value is not something that
> > callers should depend on.
> 
> Ok, I'll change it to "non-zero" here.

Thanks!
Suren Baghdasaryan June 30, 2023, 9:25 p.m. UTC | #4 
On Thu, Jun 29, 2023 at 8:47 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Thu, Jun 29, 2023 at 08:45:39PM -0700, Suren Baghdasaryan wrote:
> > On Thu, Jun 29, 2023 at 8:36 PM Matthew Wilcox <willy@infradead.org> wrote:
> > >
> > > On Thu, Jun 29, 2023 at 07:04:33PM -0700, Suren Baghdasaryan wrote:
> > > > Change folio_lock_or_retry to accept vm_fault struct and return the
> > > > vm_fault_t directly.
> > >
> > > I thought we decided to call this folio_lock_fault()?
> > >
> > > > +static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
> > > > +                                          struct vm_fault *vmf)
> > > >  {
> > > >       might_sleep();
> > > > -     return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
> > > > +     return folio_trylock(folio) ? 0 : __folio_lock_or_retry(folio, vmf);
> > >
> > > No, don't use the awful ternary operator.  The || form is used
> > > everywhere else.
> >
> > Ok, but folio_trylock() returns a boolean while folio_lock_or_retry
> > should return vm_fault_t. How exactly do you suggest changing this?
> > Something like this perhaps:
> >
> > static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
> >                                           struct vm_fault *vmf)
> > {
> >      might_sleep();
> >      if (folio_trylock(folio))
> >          return 0;
> >      return __folio_lock_or_retry(folio, mm, flags);
> > }
> >
> > ?
>
> I think the automatic casting would work, but I prefer what you've
> written here.
>
> > > >  /*
> > > >   * Return values:
> > > > - * true - folio is locked; mmap_lock is still held.
> > > > - * false - folio is not locked.
> > > > + * 0 - folio is locked.
> > > > + * VM_FAULT_RETRY - folio is not locked.
> > >
> > > I don't think we want to be so prescriptive here.  It returns non-zero
> > > if the folio is not locked.  The precise value is not something that
> > > callers should depend on.
> >
> > Ok, I'll change it to "non-zero" here.
>
> Thanks!

Posted v7 at https://lore.kernel.org/all/20230630211957.1341547-1-surenb@google.com/
diff mbox series

Patch

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 716953ee1ebd..0026a0a8277c 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -900,8 +900,7 @@  static inline bool wake_page_match(struct wait_page_queue *wait_page,
 
 void __folio_lock(struct folio *folio);
 int __folio_lock_killable(struct folio *folio);
-bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
-				unsigned int flags);
+vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
 void unlock_page(struct page *page);
 void folio_unlock(struct folio *folio);
 
@@ -1005,11 +1004,11 @@  static inline int folio_lock_killable(struct folio *folio)
  * Return value and mmap_lock implications depend on flags; see
  * __folio_lock_or_retry().
  */
-static inline bool folio_lock_or_retry(struct folio *folio,
-		struct mm_struct *mm, unsigned int flags)
+static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
+					     struct vm_fault *vmf)
 {
 	might_sleep();
-	return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
+	return folio_trylock(folio) ? 0 : __folio_lock_or_retry(folio, vmf);
 }
 
 /*
diff --git a/mm/filemap.c b/mm/filemap.c
index 9e44a49bbd74..d245bb4f7153 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1669,32 +1669,34 @@  static int __folio_lock_async(struct folio *folio, struct wait_page_queue *wait)
 
 /*
  * Return values:
- * true - folio is locked; mmap_lock is still held.
- * false - folio is not locked.
+ * 0 - folio is locked.
+ * VM_FAULT_RETRY - folio is not locked.
  *     mmap_lock has been released (mmap_read_unlock(), unless flags had both
  *     FAULT_FLAG_ALLOW_RETRY and FAULT_FLAG_RETRY_NOWAIT set, in
  *     which case mmap_lock is still held.
  *
- * If neither ALLOW_RETRY nor KILLABLE are set, will always return true
+ * If neither ALLOW_RETRY nor KILLABLE are set, will always return 0
  * with the folio locked and the mmap_lock unperturbed.
  */
-bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
-			 unsigned int flags)
+vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf)
 {
+	struct mm_struct *mm = vmf->vma->vm_mm;
+	unsigned int flags = vmf->flags;
+
 	if (fault_flag_allow_retry_first(flags)) {
 		/*
 		 * CAUTION! In this case, mmap_lock is not released
-		 * even though return 0.
+		 * even though return VM_FAULT_RETRY.
 		 */
 		if (flags & FAULT_FLAG_RETRY_NOWAIT)
-			return false;
+			return VM_FAULT_RETRY;
 
 		mmap_read_unlock(mm);
 		if (flags & FAULT_FLAG_KILLABLE)
 			folio_wait_locked_killable(folio);
 		else
 			folio_wait_locked(folio);
-		return false;
+		return VM_FAULT_RETRY;
 	}
 	if (flags & FAULT_FLAG_KILLABLE) {
 		bool ret;
@@ -1702,13 +1704,13 @@  bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
 		ret = __folio_lock_killable(folio);
 		if (ret) {
 			mmap_read_unlock(mm);
-			return false;
+			return VM_FAULT_RETRY;
 		}
 	} else {
 		__folio_lock(folio);
 	}
 
-	return true;
+	return 0;
 }
 
 /**
diff --git a/mm/memory.c b/mm/memory.c
index 5f26c56ce979..4ae3f046f593 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -3582,6 +3582,7 @@  static vm_fault_t remove_device_exclusive_entry(struct vm_fault *vmf)
 	struct folio *folio = page_folio(vmf->page);
 	struct vm_area_struct *vma = vmf->vma;
 	struct mmu_notifier_range range;
+	vm_fault_t ret;
 
 	/*
 	 * We need a reference to lock the folio because we don't hold
@@ -3594,9 +3595,10 @@  static vm_fault_t remove_device_exclusive_entry(struct vm_fault *vmf)
 	if (!folio_try_get(folio))
 		return 0;
 
-	if (!folio_lock_or_retry(folio, vma->vm_mm, vmf->flags)) {
+	ret = folio_lock_or_retry(folio, vmf);
+	if (ret) {
 		folio_put(folio);
-		return VM_FAULT_RETRY;
+		return ret;
 	}
 	mmu_notifier_range_init_owner(&range, MMU_NOTIFY_EXCLUSIVE, 0,
 				vma->vm_mm, vmf->address & PAGE_MASK,
@@ -3721,7 +3723,6 @@  vm_fault_t do_swap_page(struct vm_fault *vmf)
 	bool exclusive = false;
 	swp_entry_t entry;
 	pte_t pte;
-	int locked;
 	vm_fault_t ret = 0;
 	void *shadow = NULL;
 
@@ -3844,12 +3845,9 @@  vm_fault_t do_swap_page(struct vm_fault *vmf)
 		goto out_release;
 	}
 
-	locked = folio_lock_or_retry(folio, vma->vm_mm, vmf->flags);
-
-	if (!locked) {
-		ret |= VM_FAULT_RETRY;
+	ret |= folio_lock_or_retry(folio, vmf);
+	if (ret & VM_FAULT_RETRY)
 		goto out_release;
-	}
 
 	if (swapcache) {
 		/*