+ drivers-ibqib-optimize-mmap_sem-usage.patch added to -mm tree

[akpm@xxxxxxxxxxxxxxxxxxxx]

The patch titled
     Subject: drivers/IB,qib: optimize mmap_sem usage
has been added to the -mm tree.  Its filename is
     drivers-ibqib-optimize-mmap_sem-usage.patch

This patch should soon appear at
    http://ozlabs.org/~akpm/mmots/broken-out/drivers-ibqib-optimize-mmap_sem-usage.patch
and later at
    http://ozlabs.org/~akpm/mmotm/broken-out/drivers-ibqib-optimize-mmap_sem-usage.patch

Before you just go and hit "reply", please:
   a) Consider who else should be cc'ed
   b) Prefer to cc a suitable mailing list as well
   c) Ideally: find the original patch on the mailing list and do a
      reply-to-all to that, adding suitable additional cc's

*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***

The -mm tree is included into linux-next and is updated
there every 3-4 working days

------------------------------------------------------
From: Davidlohr Bueso <dave@xxxxxxxxxxxx>
Subject: drivers/IB,qib: optimize mmap_sem usage

The driver uses mmap_sem for both pinned_vm accounting and
get_user_pages().  Because rdma drivers might want to use gup_longterm()
in the future we still need some sort of mmap_sem serialization (as
opposed to removing it entirely by using gup_fast()).  Now that pinned_vm
is atomic the writer lock can therefore be converted to reader.

This also fixes a bug that __qib_get_user_pages was not taking into
account the current value of pinned_vm.

Link: http://lkml.kernel.org/r/20190206175920.31082-4-dave@xxxxxxxxxxxx
Signed-off-by: Davidlohr Bueso <dbueso@xxxxxxx>
Reviewed-by: Ira Weiny <ira.weiny@xxxxxxxxx>
Cc: Dennis Dalessandro <dennis.dalessandro@xxxxxxxxx>
Cc: Mike Marciniszyn <mike.marciniszyn@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---


--- a/drivers/infiniband/hw/qib/qib_user_pages.c~drivers-ibqib-optimize-mmap_sem-usage
+++ a/drivers/infiniband/hw/qib/qib_user_pages.c
@@ -49,43 +49,6 @@ static void __qib_release_user_pages(str
 	}
 }
 
-/*
- * Call with current->mm->mmap_sem held.
- */
-static int __qib_get_user_pages(unsigned long start_page, size_t num_pages,
-				struct page **p)
-{
-	unsigned long lock_limit;
-	size_t got;
-	int ret;
-
-	lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
-
-	if (num_pages > lock_limit && !capable(CAP_IPC_LOCK)) {
-		ret = -ENOMEM;
-		goto bail;
-	}
-
-	for (got = 0; got < num_pages; got += ret) {
-		ret = get_user_pages_longterm(start_page + got * PAGE_SIZE,
-					      num_pages - got,
-					      FOLL_WRITE | FOLL_FORCE,
-					      p + got, NULL);
-		if (ret < 0)
-			goto bail_release;
-	}
-
-	atomic64_add(num_pages, &current->mm->pinned_vm);
-
-	ret = 0;
-	goto bail;
-
-bail_release:
-	__qib_release_user_pages(p, got, 0);
-bail:
-	return ret;
-}
-
 /**
  * qib_map_page - a safety wrapper around pci_map_page()
  *
@@ -137,26 +100,44 @@ int qib_map_page(struct pci_dev *hwdev,
 int qib_get_user_pages(unsigned long start_page, size_t num_pages,
 		       struct page **p)
 {
+	unsigned long locked, lock_limit;
+	size_t got;
 	int ret;
 
-	down_write(&current->mm->mmap_sem);
+	lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
+	locked = atomic64_add_return(num_pages, &current->mm->pinned_vm);
 
-	ret = __qib_get_user_pages(start_page, num_pages, p);
+	if (num_pages > lock_limit && !capable(CAP_IPC_LOCK)) {
+		ret = -ENOMEM;
+		goto bail;
+	}
 
-	up_write(&current->mm->mmap_sem);
+	down_read(&current->mm->mmap_sem);
+	for (got = 0; got < num_pages; got += ret) {
+		ret = get_user_pages_longterm(start_page + got * PAGE_SIZE,
+					      num_pages - got,
+					      FOLL_WRITE | FOLL_FORCE,
+					      p + got, NULL);
+		if (ret < 0) {
+			up_read(&current->mm->mmap_sem);
+			goto bail_release;
+		}
+	}
+	up_read(&current->mm->mmap_sem);
 
+	return 0;
+bail_release:
+	__qib_release_user_pages(p, got, 0);
+bail:
+	atomic64_sub(num_pages, &current->mm->pinned_vm);
 	return ret;
 }
 
 void qib_release_user_pages(struct page **p, size_t num_pages)
 {
-	if (current->mm) /* during close after signal, mm can be NULL */
-		down_write(&current->mm->mmap_sem);
-
 	__qib_release_user_pages(p, num_pages, 1);
 
-	if (current->mm) {
+	/* during close after signal, mm can be NULL */
+	if (current->mm)
 		atomic64_sub(num_pages, &current->mm->pinned_vm);
-		up_write(&current->mm->mmap_sem);
-	}
 }
_

Patches currently in -mm which might be from dave@xxxxxxxxxxxx are

mm-make-mm-pinned_vm-an-atomic64-counter.patch
drivers-mic-scif-do-not-use-mmap_sem.patch
drivers-ibqib-optimize-mmap_sem-usage.patch
drivers-ibhfi1-do-not-se-mmap_sem.patch
drivers-ibusnic-reduce-scope-of-mmap_sem.patch
drivers-ibcore-reduce-scope-of-mmap_sem.patch