[RFC PATCH 0/4] iov_iter: Add extraction helpers

[RFC PATCH 0/4] iov_iter: Add extraction helpers

[David Howells]

Hi Al,

Here are four patches to provide support for extracting pages from an
iov_iter, where such a thing makes sense, if you could take a look?

The first couple of patches provide iov_iter general stuff:

 (1) Move the FOLL_* flags to linux/mm_types.h so that linux/uio.h can make
     use of them.

 (2) Add a function to list-only, get or pin pages from an iterator as a
     future replacement for iov_iter_get_pages*().  Pointers to the pages
     are placed into an array (which will get allocated if not provided)
     and, depending on the iterator type and direction, the pages will have
     a ref or a pin get on them, or left untouched (on the assumption that
     the caller manages their lifetime).

     The determination is:

	UBUF/IOVEC + READ	-> pin
	UBUF/IOVEC + WRITE	-> get
	PIPE + READ		-> list-only
	BVEC/XARRAY		-> list-only
	Anything else		-> EFAULT

     It also adds a function by which the caller can determine which of
     "list only, get or pin" the extraction function will actually do to
     aid in cleaning up (returning 0, FOLL_GET or FOLL_PIN as appropriate).

Then there are a couple of patches that add stuff to netfslib that I want
to use there as well as in cifs:

 (3) Add a netfslib function to use (2) to extract pages from an ITER_IOBUF
     or ITER_UBUF iterator into an ITER_BVEC iterator.  This will get or
     pin the pages as appropriate.

 (4) Add a netfslib function to extract pages from an iterator that's of
     type ITER_UBUF/IOVEC/BVEC/KVEC/XARRAY and add them to a scatterlist.

     The function in (2) is used for a UBUF and IOVEC iterators, so those
     need cleaning up afterwards.  BVEC and XARRAY iterators are ungot and
     unpinned and may be rendered into elements that span multiple pages,
     for example if large folios are present.

I've pushed the patches here also:

	https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/log/?h=iov-extract

David

Link: https://lore.kernel.org/r/166697254399.61150.1256557652599252121.stgit@warthog.procyon.org.uk/
---
David Howells (4):
      mm: Move FOLL_* defs to mm_types.h
      iov_iter: Add a function to extract a page list from an iterator
      netfs: Add a function to extract a UBUF or IOVEC into a BVEC iterator
      netfs: Add a function to extract an iterator into a scatterlist


 fs/netfs/Makefile        |   1 +
 fs/netfs/iterator.c      | 346 +++++++++++++++++++++++++++++++++++++++
 include/linux/mm.h       |  74 ---------
 include/linux/mm_types.h |  73 +++++++++
 include/linux/netfs.h    |   5 +
 include/linux/uio.h      |  29 ++++
 lib/iov_iter.c           | 333 +++++++++++++++++++++++++++++++++++++
 mm/vmalloc.c             |   1 +
 8 files changed, 788 insertions(+), 74 deletions(-)
 create mode 100644 fs/netfs/iterator.c

[RFC PATCH 1/4] mm: Move FOLL_* defs to mm_types.h

[David Howells]

Move FOLL_* definitions to linux/mm_types.h to make them more accessible
without having to drag in all of linux/mm.h and everything that drags in
too[1].

Suggested-by: Matthew Wilcox <willy@infradead.org>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: John Hubbard <jhubbard@nvidia.com>
cc: Al Viro <viro@zeniv.linux.org.uk>
cc: linux-mm@kvack.org
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/linux-fsdevel/Y1%2FhSO+7kAJhGShG@casper.infradead.org/ [1]
---

 include/linux/mm.h       |   74 ----------------------------------------------
 include/linux/mm_types.h |   73 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 73 insertions(+), 74 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 8bbcccbc5565..7a7a287818ad 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2941,80 +2941,6 @@ static inline vm_fault_t vmf_error(int err)
 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
 			 unsigned int foll_flags);
 
-#define FOLL_WRITE	0x01	/* check pte is writable */
-#define FOLL_TOUCH	0x02	/* mark page accessed */
-#define FOLL_GET	0x04	/* do get_page on page */
-#define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
-#define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
-#define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
-				 * and return without waiting upon it */
-#define FOLL_NOFAULT	0x80	/* do not fault in pages */
-#define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
-#define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
-#define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
-#define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
-#define FOLL_ANON	0x8000	/* don't do file mappings */
-#define FOLL_LONGTERM	0x10000	/* mapping lifetime is indefinite: see below */
-#define FOLL_SPLIT_PMD	0x20000	/* split huge pmd before returning */
-#define FOLL_PIN	0x40000	/* pages must be released via unpin_user_page */
-#define FOLL_FAST_ONLY	0x80000	/* gup_fast: prevent fall-back to slow gup */
-
-/*
- * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
- * other. Here is what they mean, and how to use them:
- *
- * FOLL_LONGTERM indicates that the page will be held for an indefinite time
- * period _often_ under userspace control.  This is in contrast to
- * iov_iter_get_pages(), whose usages are transient.
- *
- * FIXME: For pages which are part of a filesystem, mappings are subject to the
- * lifetime enforced by the filesystem and we need guarantees that longterm
- * users like RDMA and V4L2 only establish mappings which coordinate usage with
- * the filesystem.  Ideas for this coordination include revoking the longterm
- * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
- * added after the problem with filesystems was found FS DAX VMAs are
- * specifically failed.  Filesystem pages are still subject to bugs and use of
- * FOLL_LONGTERM should be avoided on those pages.
- *
- * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
- * Currently only get_user_pages() and get_user_pages_fast() support this flag
- * and calls to get_user_pages_[un]locked are specifically not allowed.  This
- * is due to an incompatibility with the FS DAX check and
- * FAULT_FLAG_ALLOW_RETRY.
- *
- * In the CMA case: long term pins in a CMA region would unnecessarily fragment
- * that region.  And so, CMA attempts to migrate the page before pinning, when
- * FOLL_LONGTERM is specified.
- *
- * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
- * but an additional pin counting system) will be invoked. This is intended for
- * anything that gets a page reference and then touches page data (for example,
- * Direct IO). This lets the filesystem know that some non-file-system entity is
- * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
- * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
- * a call to unpin_user_page().
- *
- * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
- * and separate refcounting mechanisms, however, and that means that each has
- * its own acquire and release mechanisms:
- *
- *     FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
- *
- *     FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
- *
- * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
- * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
- * calls applied to them, and that's perfectly OK. This is a constraint on the
- * callers, not on the pages.)
- *
- * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
- * directly by the caller. That's in order to help avoid mismatches when
- * releasing pages: get_user_pages*() pages must be released via put_page(),
- * while pin_user_pages*() pages must be released via unpin_user_page().
- *
- * Please see Documentation/core-api/pin_user_pages.rst for more information.
- */
-
 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
 {
 	if (vm_fault & VM_FAULT_OOM)
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 500e536796ca..0c80a5ad6e6a 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -1003,4 +1003,77 @@ enum fault_flag {
 
 typedef unsigned int __bitwise zap_flags_t;
 
+/*
+ * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
+ * other. Here is what they mean, and how to use them:
+ *
+ * FOLL_LONGTERM indicates that the page will be held for an indefinite time
+ * period _often_ under userspace control.  This is in contrast to
+ * iov_iter_get_pages(), whose usages are transient.
+ *
+ * FIXME: For pages which are part of a filesystem, mappings are subject to the
+ * lifetime enforced by the filesystem and we need guarantees that longterm
+ * users like RDMA and V4L2 only establish mappings which coordinate usage with
+ * the filesystem.  Ideas for this coordination include revoking the longterm
+ * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
+ * added after the problem with filesystems was found FS DAX VMAs are
+ * specifically failed.  Filesystem pages are still subject to bugs and use of
+ * FOLL_LONGTERM should be avoided on those pages.
+ *
+ * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
+ * Currently only get_user_pages() and get_user_pages_fast() support this flag
+ * and calls to get_user_pages_[un]locked are specifically not allowed.  This
+ * is due to an incompatibility with the FS DAX check and
+ * FAULT_FLAG_ALLOW_RETRY.
+ *
+ * In the CMA case: long term pins in a CMA region would unnecessarily fragment
+ * that region.  And so, CMA attempts to migrate the page before pinning, when
+ * FOLL_LONGTERM is specified.
+ *
+ * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
+ * but an additional pin counting system) will be invoked. This is intended for
+ * anything that gets a page reference and then touches page data (for example,
+ * Direct IO). This lets the filesystem know that some non-file-system entity is
+ * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
+ * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
+ * a call to unpin_user_page().
+ *
+ * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
+ * and separate refcounting mechanisms, however, and that means that each has
+ * its own acquire and release mechanisms:
+ *
+ *     FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
+ *
+ *     FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
+ *
+ * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
+ * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
+ * calls applied to them, and that's perfectly OK. This is a constraint on the
+ * callers, not on the pages.)
+ *
+ * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
+ * directly by the caller. That's in order to help avoid mismatches when
+ * releasing pages: get_user_pages*() pages must be released via put_page(),
+ * while pin_user_pages*() pages must be released via unpin_user_page().
+ *
+ * Please see Documentation/core-api/pin_user_pages.rst for more information.
+ */
+#define FOLL_WRITE	0x01	/* check pte is writable */
+#define FOLL_TOUCH	0x02	/* mark page accessed */
+#define FOLL_GET	0x04	/* do get_page on page */
+#define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
+#define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
+#define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
+				 * and return without waiting upon it */
+#define FOLL_NOFAULT	0x80	/* do not fault in pages */
+#define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
+#define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
+#define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
+#define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
+#define FOLL_ANON	0x8000	/* don't do file mappings */
+#define FOLL_LONGTERM	0x10000	/* mapping lifetime is indefinite: see below */
+#define FOLL_SPLIT_PMD	0x20000	/* split huge pmd before returning */
+#define FOLL_PIN	0x40000	/* pages must be released via unpin_user_page */
+#define FOLL_FAST_ONLY	0x80000	/* gup_fast: prevent fall-back to slow gup */
+
 #endif /* _LINUX_MM_TYPES_H */

[RFC PATCH 2/4] iov_iter: Add a function to extract a page list from an iterator

[David Howells]

Add a function, iov_iter_extract_pages(), to extract a list of pages from
an iterator.  The pages may be returned with a reference added or a pin
added or neither, depending on the type of iterator and the direction of
transfer.

An additional function, iov_iter_extract_mode() is also provided so that the
mode of retention that will be employed for an iterator can be queried - and
therefore how the caller should dispose of the pages later.

There are three cases:

 (1) Transfer *into* an ITER_IOVEC or ITER_UBUF iterator.

     Extracted pages will have pins obtained on them (but not references)
     so that fork() doesn't CoW the pages incorrectly whilst the I/O is in
     progress.

     iov_iter_extract_mode() will return FOLL_PIN for this case.  The caller
     should use something like unpin_user_page() to dispose of the page.

 (2) Transfer is *out of* an ITER_IOVEC or ITER_UBUF iterator.

     Extracted pages will have references obtained on them, but not pins.

     iov_iter_extract_mode() will return FOLL_GET.  The caller should use
     something like put_page() for page disposal.

 (3) Any other sort of iterator.

     No refs or pins are obtained on the page, the assumption is made that
     the caller will manage page retention.

     iov_iter_extract_mode() will return 0.  The pages don't need additional
     disposal.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Al Viro <viro@zeniv.linux.org.uk>
cc: Christoph Hellwig <hch@lst.de>
cc: John Hubbard <jhubbard@nvidia.com>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-fsdevel@vger.kernel.org
cc: linux-mm@kvack.org
Link: https://lore.kernel.org/r/166722777971.2555743.12953624861046741424.stgit@warthog.procyon.org.uk/
---

 include/linux/uio.h |   29 ++++
 lib/iov_iter.c      |  333 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 362 insertions(+)

diff --git a/include/linux/uio.h b/include/linux/uio.h
index 2e3134b14ffd..329e36d41f0a 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -351,4 +351,33 @@ static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction,
 	};
 }
 
+ssize_t iov_iter_extract_pages(struct iov_iter *i, struct page ***pages,
+			       size_t maxsize, unsigned int maxpages,
+			       size_t *offset0);
+
+/**
+ * iov_iter_extract_mode - Indicate how pages from the iterator will be retained
+ * @iter: The iterator
+ *
+ * Examine the indicator and indicate with FOLL_PIN, FOLL_GET or 0 as to how,
+ * if at all, pages extracted from the iterator will be retained by the
+ * extraction function.
+ *
+ * FOLL_GET indicates that the pages will have a reference taken on them that
+ * the caller must put.  This can be done for DMA/async DIO write from a page.
+ *
+ * FOLL_PIN indicates that the pages will have a pin placed in them that the
+ * caller must unpin.  This is must be done for DMA/async DIO read to a page to
+ * avoid CoW problems in fork.
+ *
+ * 0 indicates that no measures are taken and that it's up to the caller to
+ * retain the pages.
+ */
+static inline unsigned int iov_iter_extract_mode(struct iov_iter *iter)
+{
+	if (user_backed_iter(iter))
+		return iter->data_source ? FOLL_GET : FOLL_PIN;
+	return 0;
+}
+
 #endif
diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index c3ca28ca68a6..17f63f4d499b 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -1892,3 +1892,336 @@ void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state)
 		i->iov -= state->nr_segs - i->nr_segs;
 	i->nr_segs = state->nr_segs;
 }
+
+/*
+ * Extract a list of contiguous pages from an ITER_PIPE iterator.  This does
+ * not get references of its own on the pages, nor does it get a pin on them.
+ * If there's a partial page, it adds that first and will then allocate and add
+ * pages into the pipe to make up the buffer space to the amount required.
+ *
+ * The caller must hold the pipe locked and only transferring into a pipe is
+ * supported.
+ */
+static ssize_t iov_iter_extract_pipe_pages(struct iov_iter *i,
+					   struct page ***pages, size_t maxsize,
+					   unsigned int maxpages,
+					   size_t *offset0)
+{
+	unsigned int nr, offset, chunk, j;
+	struct page **p;
+	size_t left;
+
+	if (!sanity(i))
+		return -EFAULT;
+
+	offset = pipe_npages(i, &nr);
+	if (!nr)
+		return -EFAULT;
+	*offset0 = offset;
+
+	maxpages = min_t(size_t, nr, maxpages);
+	maxpages = want_pages_array(pages, maxsize, offset, maxpages);
+	if (!maxpages)
+		return -ENOMEM;
+	p = *pages;
+
+	left = maxsize;
+	for (j = 0; j < maxpages; j++) {
+		struct page *page = append_pipe(i, left, &offset);
+		if (!page)
+			break;
+		chunk = min_t(size_t, left, PAGE_SIZE - offset);
+		left -= chunk;
+		*p++ = page;
+	}
+	if (!j)
+		return -EFAULT;
+	return maxsize - left;
+}
+
+/*
+ * Extract a list of contiguous pages from an ITER_XARRAY iterator.  This does not
+ * get references on the pages, nor does it get a pin on them.
+ */
+static ssize_t iov_iter_extract_xarray_pages(struct iov_iter *i,
+					     struct page ***pages, size_t maxsize,
+					     unsigned int maxpages,
+					     size_t *offset0)
+{
+	struct page *page, **p;
+	unsigned int nr = 0, offset;
+	loff_t pos = i->xarray_start + i->iov_offset;
+	pgoff_t index = pos >> PAGE_SHIFT;
+	XA_STATE(xas, i->xarray, index);
+
+	offset = pos & ~PAGE_MASK;
+	*offset0 = offset;
+
+	maxpages = want_pages_array(pages, maxsize, offset, maxpages);
+	if (!maxpages)
+		return -ENOMEM;
+	p = *pages;
+
+	rcu_read_lock();
+	for (page = xas_load(&xas); page; page = xas_next(&xas)) {
+		if (xas_retry(&xas, page))
+			continue;
+
+		/* Has the page moved or been split? */
+		if (unlikely(page != xas_reload(&xas))) {
+			xas_reset(&xas);
+			continue;
+		}
+
+		p[nr++] = find_subpage(page, xas.xa_index);
+		if (nr == maxpages)
+			break;
+	}
+	rcu_read_unlock();
+
+	maxsize = min_t(size_t, nr * PAGE_SIZE - offset, maxsize);
+	i->iov_offset += maxsize;
+	i->count -= maxsize;
+	return maxsize;
+}
+
+/*
+ * Extract a list of contiguous pages from an ITER_BVEC iterator.  This does
+ * not get references on the pages, nor does it get a pin on them.
+ */
+static ssize_t iov_iter_extract_bvec_pages(struct iov_iter *i,
+					   struct page ***pages, size_t maxsize,
+					   unsigned int maxpages,
+					   size_t *offset0)
+{
+	struct page **p, *page;
+	size_t skip = i->iov_offset, offset;
+	int k;
+
+	maxsize = min(maxsize, i->bvec->bv_len - skip);
+	skip += i->bvec->bv_offset;
+	page = i->bvec->bv_page + skip / PAGE_SIZE;
+	offset = skip % PAGE_SIZE;
+	*offset0 = offset;
+
+	maxpages = want_pages_array(pages, maxsize, offset, maxpages);
+	if (!maxpages)
+		return -ENOMEM;
+	p = *pages;
+	for (k = 0; k < maxpages; k++)
+		p[k] = page + k;
+
+	maxsize = min_t(size_t, maxsize, maxpages * PAGE_SIZE - offset);
+	i->count -= maxsize;
+	i->iov_offset += maxsize;
+	if (i->iov_offset == i->bvec->bv_len) {
+		i->iov_offset = 0;
+		i->bvec++;
+		i->nr_segs--;
+	}
+	return maxsize;
+}
+
+/*
+ * Get the first segment from an ITER_UBUF or ITER_IOVEC iterator.  The
+ * iterator must not be empty.
+ */
+static unsigned long iov_iter_extract_first_user_segment(const struct iov_iter *i,
+							 size_t *size)
+{
+	size_t skip;
+	long k;
+
+	if (iter_is_ubuf(i))
+		return (unsigned long)i->ubuf + i->iov_offset;
+
+	for (k = 0, skip = i->iov_offset; k < i->nr_segs; k++, skip = 0) {
+		size_t len = i->iov[k].iov_len - skip;
+
+		if (unlikely(!len))
+			continue;
+		if (*size > len)
+			*size = len;
+		return (unsigned long)i->iov[k].iov_base + skip;
+	}
+	BUG(); // if it had been empty, we wouldn't get called
+}
+
+/*
+ * Extract a list of contiguous pages from a user iterator and get references
+ * on them.  This should only be used iff the iterator is user-backed
+ * (IOBUF/UBUF) and data is being transferred out of the buffer described by
+ * the iterator (ie. this is the source).
+ *
+ * The pages are returned with incremented refcounts that the caller must undo
+ * once the transfer is complete, but no additional pins are obtained.
+ *
+ * This is only safe to be used where background IO/DMA is not going to be
+ * modifying the buffer, and so won't cause a problem with CoW on fork.
+ */
+static ssize_t iov_iter_extract_user_pages_and_get(struct iov_iter *i,
+						   struct page ***pages,
+						   size_t maxsize,
+						   unsigned int maxpages,
+						   size_t *offset0)
+{
+	unsigned long addr;
+	unsigned int gup_flags = FOLL_GET;
+	size_t offset;
+	int res;
+
+	if (WARN_ON_ONCE(iov_iter_rw(i) != WRITE))
+		return -EFAULT;
+
+	if (i->nofault)
+		gup_flags |= FOLL_NOFAULT;
+
+	addr = iov_iter_extract_first_user_segment(i, &maxsize);
+	*offset0 = offset = addr % PAGE_SIZE;
+	addr &= PAGE_MASK;
+	maxpages = want_pages_array(pages, maxsize, offset, maxpages);
+	if (!maxpages)
+		return -ENOMEM;
+	res = get_user_pages_fast(addr, maxpages, gup_flags, *pages);
+	if (unlikely(res <= 0))
+		return res;
+	maxsize = min_t(size_t, maxsize, res * PAGE_SIZE - offset);
+	iov_iter_advance(i, maxsize);
+	return maxsize;
+}
+
+/*
+ * Extract a list of contiguous pages from a user iterator and get a pin on
+ * each of them.  This should only be used iff the iterator is user-backed
+ * (IOBUF/UBUF) and data is being transferred into the buffer described by the
+ * iterator (ie. this is the destination).
+ *
+ * It does not get refs on the pages, but the pages must be unpinned by the
+ * caller once the transfer is complete.
+ *
+ * This is safe to be used where background IO/DMA *is* going to be modifying
+ * the buffer; using a pin rather than a ref makes sure that CoW happens
+ * correctly in the parent during fork.
+ */
+static ssize_t iov_iter_extract_user_pages_and_pin(struct iov_iter *i,
+						   struct page ***pages,
+						   size_t maxsize,
+						   unsigned int maxpages,
+						   size_t *offset0)
+{
+	unsigned long addr;
+	unsigned int gup_flags = FOLL_PIN | FOLL_WRITE;
+	size_t offset;
+	int res;
+
+	if (WARN_ON_ONCE(iov_iter_rw(i) != READ))
+		return -EFAULT;
+
+	if (i->nofault)
+		gup_flags |= FOLL_NOFAULT;
+
+	addr = first_iovec_segment(i, &maxsize);
+	*offset0 = offset = addr % PAGE_SIZE;
+	addr &= PAGE_MASK;
+	maxpages = want_pages_array(pages, maxsize, offset, maxpages);
+	if (!maxpages)
+		return -ENOMEM;
+	res = pin_user_pages_fast(addr, maxpages, gup_flags, *pages);
+	if (unlikely(res <= 0))
+		return res;
+	maxsize = min_t(size_t, maxsize, res * PAGE_SIZE - offset);
+	iov_iter_advance(i, maxsize);
+	return maxsize;
+}
+
+static ssize_t iov_iter_extract_user_pages(struct iov_iter *i,
+					   struct page ***pages, size_t maxsize,
+					   unsigned int maxpages,
+					   size_t *offset0)
+{
+	switch (iov_iter_extract_mode(i)) {
+	case FOLL_GET:
+		return iov_iter_extract_user_pages_and_get(i, pages, maxsize,
+							   maxpages, offset0);
+	case FOLL_PIN:
+		return iov_iter_extract_user_pages_and_pin(i, pages, maxsize,
+							   maxpages, offset0);
+	default:
+		BUG();
+	}
+}
+
+/**
+ * iov_iter_extract_pages - Extract a list of contiguous pages from an iterator
+ * @i: The iterator to extract from
+ * @pages: Where to return the list of pages
+ * @maxsize: The maximum amount of iterator to extract
+ * @maxpages: The maximum size of the list of pages
+ * @offset0: Where to return the starting offset into (*@pages)[0]
+ *
+ * Extract a list of contiguous pages from the current point of the iterator,
+ * advancing the iterator.  The maximum number of pages and the maximum amount
+ * of page contents can be set.
+ *
+ * If *@pages is NULL, a page list will be allocated to the required size and
+ * *@pages will be set to its base.  If *@pages is not NULL, it will be assumed
+ * that the caller allocated a page list at least @maxpages in size and this
+ * will be filled in.
+ *
+ * Extra refs or pins on the pages may be obtained as follows:
+ *
+ *  (*) If the iterator is user-backed (ITER_IOVEC/ITER_UBUF) and data is to be
+ *      transferred /OUT OF/ the described buffer, refs will be taken on the
+ *      pages, but pins will not be added.  This can be used for DMA from a
+ *      page; it cannot be used for DMA to a page, as it may cause page-COW
+ *      problems in fork.
+ *
+ *  (*) If the iterator is user-backed (ITER_IOVEC/ITER_UBUF) and data is to be
+ *      transferred /INTO/ the described buffer, pins will be added to the
+ *      pages, but refs will not be taken.  This must be used for DMA to a
+ *      page.
+ *
+ *  (*) If the iterator is ITER_PIPE, this must describe a destination for the
+ *      data.  Additional pages may be allocated and added to the pipe (which
+ *      will hold the refs), but neither refs nor pins will be obtained for the
+ *      caller.  The caller must hold the pipe lock.
+ *
+ *  (*) If the iterator is ITER_BVEC or ITER_XARRAY, the pages are merely
+ *      listed; no extra refs or pins are obtained.
+ *
+ * Note also:
+ *
+ *  (*) Use with ITER_KVEC is not supported as that may refer to memory that
+ *      doesn't have associated page structs.
+ *
+ *  (*) Use with ITER_DISCARD is not supported as that has no content.
+ *
+ * On success, the function sets *@pages to the new pagelist, if allocated, and
+ * sets *offset0 to the offset into the first page and returns the amount of
+ * buffer space added represented by the page list.
+ *
+ * It may also return -ENOMEM and -EFAULT.
+ */
+ssize_t iov_iter_extract_pages(struct iov_iter *i, struct page ***pages,
+			       size_t maxsize, unsigned int maxpages,
+			       size_t *offset0)
+{
+	maxsize = min_t(size_t, min_t(size_t, maxsize, i->count), MAX_RW_COUNT);
+	if (!maxsize)
+		return 0;
+
+	if (likely(user_backed_iter(i)))
+		return iov_iter_extract_user_pages(i, pages, maxsize,
+						   maxpages, offset0);
+	if (iov_iter_is_bvec(i))
+		return iov_iter_extract_bvec_pages(i, pages, maxsize,
+						   maxpages, offset0);
+	if (iov_iter_is_pipe(i))
+		return iov_iter_extract_pipe_pages(i, pages, maxsize,
+						   maxpages, offset0);
+	if (iov_iter_is_xarray(i))
+		return iov_iter_extract_xarray_pages(i, pages, maxsize,
+						     maxpages, offset0);
+	return -EFAULT;
+}
+EXPORT_SYMBOL(iov_iter_extract_pages);

[RFC PATCH 3/4] netfs: Add a function to extract a UBUF or IOVEC into a BVEC iterator

[David Howells]

Add a function to extract the pages from a user-space supplied iterator
(UBUF- or IOVEC-type) into a BVEC-type iterator, pinning the pages as we
go.

This is useful in three situations:

 (1) A userspace thread may have a sibling that unmaps or remaps the
     process's VM during the operation, changing the assignment of the
     pages and potentially causing an error.  Pinning the pages keeps
     some pages around, even if this occurs; futher, we find out at the
     point of extraction if EFAULT is going to be incurred.

 (2) Pages might get swapped out/discarded if not pinned, so we want to pin
     them to avoid the reload causing a deadlock due to a DIO from/to an
     mmapped region on the same file.

 (3) The iterator may get passed to sendmsg() by the filesystem.  If a
     fault occurs, we may get a short write to a TCP stream that's then
     tricky to recover from.

We assume that other types of iterator (eg. BVEC-, KVEC- and XARRAY-type)
are constructed only by kernel internals and that the pages are pinned in
those cases.

DISCARD- and PIPE-type iterators aren't DIO'able.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Steve French <sfrench@samba.org>
cc: Shyam Prasad N <nspmangalore@gmail.com>
cc: Rohith Surabattula <rohiths.msft@gmail.com>
cc: linux-cachefs@redhat.com
cc: linux-cifs@vger.kernel.org
cc: linux-fsdevel@vger.kernel.org
---

 fs/netfs/Makefile     |    1 +
 fs/netfs/iterator.c   |   94 +++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/netfs.h |    2 +
 3 files changed, 97 insertions(+)
 create mode 100644 fs/netfs/iterator.c

diff --git a/fs/netfs/Makefile b/fs/netfs/Makefile
index f684c0cd1ec5..386d6fb92793 100644
--- a/fs/netfs/Makefile
+++ b/fs/netfs/Makefile
@@ -3,6 +3,7 @@
 netfs-y := \
 	buffered_read.o \
 	io.o \
+	iterator.o \
 	main.o \
 	objects.o
 
diff --git a/fs/netfs/iterator.c b/fs/netfs/iterator.c
new file mode 100644
index 000000000000..c11d05a66a4a
--- /dev/null
+++ b/fs/netfs/iterator.c
@@ -0,0 +1,94 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* Iterator helpers.
+ *
+ * Copyright (C) 2022 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#include <linux/export.h>
+#include <linux/slab.h>
+#include <linux/uio.h>
+#include <linux/netfs.h>
+#include "internal.h"
+
+/**
+ * netfs_extract_user_iter - Extract the pages from a user iterator into a bvec
+ * @orig: The original iterator
+ * @orig_len: The amount of iterator to copy
+ * @new: The iterator to be set up
+ *
+ * Extract the page fragments from the given amount of the source iterator and
+ * build up a second iterator that refers to all of those bits.  This allows
+ * the original iterator to disposed of.
+ *
+ * On success, the number of elements in the bvec is returned and the original
+ * iterator will have been advanced by the amount extracted.
+ */
+ssize_t netfs_extract_user_iter(struct iov_iter *orig, size_t orig_len,
+				struct iov_iter *new)
+{
+	struct bio_vec *bv = NULL;
+	struct page **pages;
+	unsigned int cur_npages;
+	unsigned int max_pages;
+	unsigned int npages = 0;
+	unsigned int i;
+	ssize_t ret;
+	size_t count = orig_len, offset, len;
+	size_t bv_size, pg_size;
+
+	if (WARN_ON_ONCE(!iter_is_ubuf(orig) && !iter_is_iovec(orig)))
+		return -EIO;
+
+	max_pages = iov_iter_npages(orig, INT_MAX);
+	bv_size = array_size(max_pages, sizeof(*bv));
+	bv = kvmalloc(bv_size, GFP_KERNEL);
+	if (!bv)
+		return -ENOMEM;
+
+	/* Put the page list at the end of the bvec list storage.  bvec
+	 * elements are larger than page pointers, so as long as we work
+	 * 0->last, we should be fine.
+	 */
+	pg_size = array_size(max_pages, sizeof(*pages));
+	pages = (void *)bv + bv_size - pg_size;
+
+	while (count && npages < max_pages) {
+		ret = iov_iter_extract_pages(orig, &pages, count,
+					     max_pages - npages, &offset);
+		if (ret < 0) {
+			pr_err("Couldn't get user pages (rc=%zd)\n", ret);
+			break;
+		}
+
+		if (ret > count) {
+			pr_err("get_pages rc=%zd more than %zu\n", ret, count);
+			break;
+		}
+
+		count -= ret;
+		ret += offset;
+		cur_npages = DIV_ROUND_UP(ret, PAGE_SIZE);
+
+		if (npages + cur_npages > max_pages) {
+			pr_err("Out of bvec array capacity (%u vs %u)\n",
+			       npages + cur_npages, max_pages);
+			break;
+		}
+
+		for (i = 0; i < cur_npages; i++) {
+			len = ret > PAGE_SIZE ? PAGE_SIZE : ret;
+			bv[npages + i].bv_page	 = *pages++;
+			bv[npages + i].bv_offset = offset;
+			bv[npages + i].bv_len	 = len - offset;
+			ret -= len;
+			offset = 0;
+		}
+
+		npages += cur_npages;
+	}
+
+	iov_iter_bvec(new, iov_iter_rw(orig), bv, npages, orig_len - count);
+	return npages;
+}
+EXPORT_SYMBOL(netfs_extract_user_iter);
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index f2402ddeafbf..5f6ad0246946 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -288,6 +288,8 @@ void netfs_get_subrequest(struct netfs_io_subrequest *subreq,
 void netfs_put_subrequest(struct netfs_io_subrequest *subreq,
 			  bool was_async, enum netfs_sreq_ref_trace what);
 void netfs_stats_show(struct seq_file *);
+ssize_t netfs_extract_user_iter(struct iov_iter *orig, size_t orig_len,
+				struct iov_iter *new);
 
 /**
  * netfs_inode - Get the netfs inode context from the inode

[RFC PATCH 4/4] netfs: Add a function to extract an iterator into a scatterlist

[David Howells]

Provide a function for filling in a scatterlist from the list of pages
contained in an iterator.

If the iterator is UBUF- or IOBUF-type, the pages have a ref taken on them.

If the iterator is BVEC-, KVEC- or XARRAY-type, no ref is taken on the
pages and it is left to the caller to manage their lifetime.  It cannot be
assumed that a ref can be validly taken, particularly in the case of a KVEC
iterator.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Steve French <sfrench@samba.org>
cc: Shyam Prasad N <nspmangalore@gmail.com>
cc: Rohith Surabattula <rohiths.msft@gmail.com>
cc: linux-cachefs@redhat.com
cc: linux-cifs@vger.kernel.org
cc: linux-fsdevel@vger.kernel.org
---

 fs/netfs/iterator.c   |  252 +++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/netfs.h |    3 +
 mm/vmalloc.c          |    1 
 3 files changed, 256 insertions(+)

diff --git a/fs/netfs/iterator.c b/fs/netfs/iterator.c
index c11d05a66a4a..62485416cc3d 100644
--- a/fs/netfs/iterator.c
+++ b/fs/netfs/iterator.c
@@ -7,7 +7,9 @@
 
 #include <linux/export.h>
 #include <linux/slab.h>
+#include <linux/mm.h>
 #include <linux/uio.h>
+#include <linux/scatterlist.h>
 #include <linux/netfs.h>
 #include "internal.h"
 
@@ -92,3 +94,253 @@ ssize_t netfs_extract_user_iter(struct iov_iter *orig, size_t orig_len,
 	return npages;
 }
 EXPORT_SYMBOL(netfs_extract_user_iter);
+
+/*
+ * Extract and pin up to sg_max pages from UBUF- or IOVEC-class iterators and
+ * add them to the scatterlist.
+ */
+static ssize_t netfs_extract_user_to_sg(struct iov_iter *iter,
+					ssize_t maxsize,
+					struct sg_table *sgtable,
+					unsigned int sg_max)
+{
+	struct scatterlist *sg = sgtable->sgl + sgtable->nents;
+	struct page **pages;
+	unsigned int npages;
+	ssize_t ret = 0, res;
+	size_t len, off;
+
+	/* We decant the page list into the tail of the scatterlist */
+	pages = (void *)sgtable->sgl + array_size(sg_max, sizeof(struct scatterlist));
+	pages -= sg_max;
+
+	do {
+		res = iov_iter_get_pages2(iter, pages, maxsize, sg_max, &off);
+		if (res < 0)
+			goto failed;
+
+		len = res;
+		maxsize -= len;
+		ret += len;
+		npages = DIV_ROUND_UP(off + len, PAGE_SIZE);
+		sg_max -= npages;
+
+		for (; npages < 0; npages--) {
+			struct page *page = *pages;
+			size_t seg = min_t(size_t, PAGE_SIZE - off, len);
+
+			*pages++ = NULL;
+			sg_set_page(sg, page, len, off);
+			sgtable->nents++;
+			sg++;
+			len -= seg;
+			off = 0;
+		}
+	} while (maxsize > 0 && sg_max > 0);
+
+	return ret;
+
+failed:
+	while (sgtable->nents > sgtable->orig_nents)
+		put_page(sg_page(&sgtable->sgl[--sgtable->nents]));
+	return res;
+}
+
+/*
+ * Extract up to sg_max pages from a BVEC-type iterator and add them to the
+ * scatterlist.  The pages are not pinned.
+ */
+static ssize_t netfs_extract_bvec_to_sg(struct iov_iter *iter,
+					ssize_t maxsize,
+					struct sg_table *sgtable,
+					unsigned int sg_max)
+{
+	const struct bio_vec *bv = iter->bvec;
+	struct scatterlist *sg = sgtable->sgl + sgtable->nents;
+	unsigned long start = iter->iov_offset;
+	unsigned int i;
+	ssize_t ret = 0;
+
+	for (i = 0; i < iter->nr_segs; i++) {
+		size_t off, len;
+
+		len = bv[i].bv_len;
+		if (start >= len) {
+			start -= len;
+			continue;
+		}
+
+		len = min_t(size_t, maxsize, len - start);
+		off = bv[i].bv_offset + start;
+
+		sg_set_page(sg, bv[i].bv_page, len, off);
+		sgtable->nents++;
+		sg++;
+		sg_max--;
+
+		ret += len;
+		maxsize -= len;
+		if (maxsize <= 0 || sg_max == 0)
+			break;
+		start = 0;
+	}
+
+	if (ret > 0)
+		iov_iter_advance(iter, ret);
+	return ret;
+}
+
+/*
+ * Extract up to sg_max pages from a KVEC-type iterator and add them to the
+ * scatterlist.  This can deal with vmalloc'd buffers as well as kmalloc'd or
+ * static buffers.  The pages are not pinned.
+ */
+static ssize_t netfs_extract_kvec_to_sg(struct iov_iter *iter,
+					ssize_t maxsize,
+					struct sg_table *sgtable,
+					unsigned int sg_max)
+{
+	const struct kvec *kv = iter->kvec;
+	struct scatterlist *sg = sgtable->sgl + sgtable->nents;
+	unsigned long start = iter->iov_offset;
+	unsigned int i;
+	ssize_t ret = 0;
+
+	for (i = 0; i < iter->nr_segs; i++) {
+		struct page *page;
+		unsigned long kaddr;
+		size_t off, len, seg;
+
+		len = kv[i].iov_len;
+		if (start >= len) {
+			start -= len;
+			continue;
+		}
+
+		kaddr = (unsigned long)kv[i].iov_base + start;
+		off = kaddr & ~PAGE_MASK;
+		len = min_t(size_t, maxsize, len - start);
+		kaddr &= PAGE_MASK;
+
+		maxsize -= len;
+		ret += len;
+		do {
+			seg = min_t(size_t, len, PAGE_SIZE - off);
+			if (is_vmalloc_or_module_addr((void *)kaddr))
+				page = vmalloc_to_page((void *)kaddr);
+			else
+				page = virt_to_page(kaddr);
+
+			sg_set_page(sg, page, len, off);
+			sgtable->nents++;
+			sg++;
+			sg_max--;
+
+			len -= seg;
+			kaddr += PAGE_SIZE;
+			off = 0;
+		} while (len > 0 && sg_max > 0);
+
+		if (maxsize <= 0 || sg_max == 0)
+			break;
+		start = 0;
+	}
+
+	if (ret > 0)
+		iov_iter_advance(iter, ret);
+	return ret;
+}
+
+/*
+ * Extract up to sg_max folios from an XARRAY-type iterator and add them to
+ * the scatterlist.  The pages are not pinned.
+ */
+static ssize_t netfs_extract_xarray_to_sg(struct iov_iter *iter,
+					  ssize_t maxsize,
+					  struct sg_table *sgtable,
+					  unsigned int sg_max)
+{
+	struct scatterlist *sg = sgtable->sgl + sgtable->nents;
+	struct xarray *xa = iter->xarray;
+	struct folio *folio;
+	loff_t start = iter->xarray_start + iter->iov_offset;
+	pgoff_t index = start / PAGE_SIZE;
+	ssize_t ret = 0;
+	size_t offset, len;
+	XA_STATE(xas, xa, index);
+
+	rcu_read_lock();
+
+	xas_for_each(&xas, folio, ULONG_MAX) {
+		if (xas_retry(&xas, folio))
+			continue;
+		if (WARN_ON(xa_is_value(folio)))
+			break;
+		if (WARN_ON(folio_test_hugetlb(folio)))
+			break;
+
+		offset = offset_in_folio(folio, start);
+		len = min_t(size_t, maxsize, folio_size(folio) - offset);
+
+		sg_set_page(sg, folio_page(folio, 0), len, offset);
+		sgtable->nents++;
+		sg++;
+		sg_max--;
+
+		maxsize -= len;
+		ret += len;
+		if (maxsize <= 0 || sg_max == 0)
+			break;
+	}
+
+	rcu_read_unlock();
+	if (ret > 0)
+		iov_iter_advance(iter, ret);
+	return ret;
+}
+
+/**
+ * netfs_extract_iter_to_sg - Extract pages from an iterator and add ot an sglist
+ * @iter: The iterator to extract from
+ * @maxsize: The amount of iterator to copy
+ * @sgtable: The scatterlist table to fill in
+ * @sg_max: Maximum number of elements in @sgtable that may be filled
+ *
+ * Extract the page fragments from the given amount of the source iterator and
+ * add them to a scatterlist that refers to all of those bits, to a maximum
+ * addition of @sg_max elements.
+ *
+ * The pages referred to by UBUF- and IOVEC-type iterators are extracted and
+ * pinned; BVEC-, KVEC- and XARRAY-type are extracted but aren't pinned; PIPE-
+ * and DISCARD-type are not supported.
+ *
+ * No end mark is placed on the scatterlist; that's left to the caller.
+ *
+ * If successul, @sgtable->nents is updated to include the number of elements
+ * added and the number of bytes added is returned.  @sgtable->orig_nents is
+ * left unaltered.
+ */
+ssize_t netfs_extract_iter_to_sg(struct iov_iter *iter, size_t maxsize,
+				 struct sg_table *sgtable, unsigned int sg_max)
+{
+	if (maxsize == 0)
+		return 0;
+
+	switch (iov_iter_type(iter)) {
+	case ITER_UBUF:
+	case ITER_IOVEC:
+		return netfs_extract_user_to_sg(iter, maxsize, sgtable, sg_max);
+	case ITER_BVEC:
+		return netfs_extract_bvec_to_sg(iter, maxsize, sgtable, sg_max);
+	case ITER_KVEC:
+		return netfs_extract_kvec_to_sg(iter, maxsize, sgtable, sg_max);
+	case ITER_XARRAY:
+		return netfs_extract_xarray_to_sg(iter, maxsize, sgtable, sg_max);
+	default:
+		pr_err("netfs_extract_iter_to_sg(%u) unsupported\n",
+		       iov_iter_type(iter));
+		WARN_ON_ONCE(1);
+		return -EIO;
+	}
+}
+EXPORT_SYMBOL(netfs_extract_iter_to_sg);
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index 5f6ad0246946..21771dd594a1 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -290,6 +290,9 @@ void netfs_put_subrequest(struct netfs_io_subrequest *subreq,
 void netfs_stats_show(struct seq_file *);
 ssize_t netfs_extract_user_iter(struct iov_iter *orig, size_t orig_len,
 				struct iov_iter *new);
+struct sg_table;
+ssize_t netfs_extract_iter_to_sg(struct iov_iter *iter, size_t len,
+				 struct sg_table *sgtable, unsigned int sg_max);
 
 /**
  * netfs_inode - Get the netfs inode context from the inode
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index ccaa461998f3..b13ac142685b 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -653,6 +653,7 @@ int is_vmalloc_or_module_addr(const void *x)
 #endif
 	return is_vmalloc_addr(x);
 }
+EXPORT_SYMBOL_GPL(is_vmalloc_or_module_addr);
 
 /*
  * Walk a vmap address to the struct page it maps. Huge vmap mappings will

Re: [RFC PATCH 1/4] mm: Move FOLL_* defs to mm_types.h

[John Hubbard]

On 11/17/22 06:54, David Howells wrote:
> Move FOLL_* definitions to linux/mm_types.h to make them more accessible
> without having to drag in all of linux/mm.h and everything that drags in
> too[1].
> 
> Suggested-by: Matthew Wilcox <willy@infradead.org>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: John Hubbard <jhubbard@nvidia.com>
> cc: Al Viro <viro@zeniv.linux.org.uk>
> cc: linux-mm@kvack.org
> cc: linux-fsdevel@vger.kernel.org
> Link: https://lore.kernel.org/linux-fsdevel/Y1%2FhSO+7kAJhGShG@casper.infradead.org/ [1]
> ---
> 
>   include/linux/mm.h       |   74 ----------------------------------------------
>   include/linux/mm_types.h |   73 +++++++++++++++++++++++++++++++++++++++++++++
>   2 files changed, 73 insertions(+), 74 deletions(-)


OK, I've verified that this is a "mostly identical" movement: the only
thing that changes is that the comments now come before the defines.

And because mm.h includes mm_types.h, it is unlikely that moving a
define from mm.h to mm_types.h would cause build failures. It's not
completely impossible: ordering issues are sometimes involved in this
sort of change. But unlikely.

Anyway, this is a good move. The users of various mm APIs should not
have to pull in quite so much of the internals of mm, and this is a step
in that direction. FOLL_* items are used by filesystems and other
subsystems that definitely do not need all of mm.h.


Reviewed-by: John Hubbard <jhubbard@nvidia.com>


thanks,
-- 
John Hubbard
NVIDIA

> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 8bbcccbc5565..7a7a287818ad 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2941,80 +2941,6 @@ static inline vm_fault_t vmf_error(int err)
>   struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
>   			 unsigned int foll_flags);
>   
> -#define FOLL_WRITE	0x01	/* check pte is writable */
> -#define FOLL_TOUCH	0x02	/* mark page accessed */
> -#define FOLL_GET	0x04	/* do get_page on page */
> -#define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
> -#define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
> -#define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
> -				 * and return without waiting upon it */
> -#define FOLL_NOFAULT	0x80	/* do not fault in pages */
> -#define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
> -#define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
> -#define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
> -#define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
> -#define FOLL_ANON	0x8000	/* don't do file mappings */
> -#define FOLL_LONGTERM	0x10000	/* mapping lifetime is indefinite: see below */
> -#define FOLL_SPLIT_PMD	0x20000	/* split huge pmd before returning */
> -#define FOLL_PIN	0x40000	/* pages must be released via unpin_user_page */
> -#define FOLL_FAST_ONLY	0x80000	/* gup_fast: prevent fall-back to slow gup */
> -
> -/*
> - * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
> - * other. Here is what they mean, and how to use them:
> - *
> - * FOLL_LONGTERM indicates that the page will be held for an indefinite time
> - * period _often_ under userspace control.  This is in contrast to
> - * iov_iter_get_pages(), whose usages are transient.
> - *
> - * FIXME: For pages which are part of a filesystem, mappings are subject to the
> - * lifetime enforced by the filesystem and we need guarantees that longterm
> - * users like RDMA and V4L2 only establish mappings which coordinate usage with
> - * the filesystem.  Ideas for this coordination include revoking the longterm
> - * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
> - * added after the problem with filesystems was found FS DAX VMAs are
> - * specifically failed.  Filesystem pages are still subject to bugs and use of
> - * FOLL_LONGTERM should be avoided on those pages.
> - *
> - * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
> - * Currently only get_user_pages() and get_user_pages_fast() support this flag
> - * and calls to get_user_pages_[un]locked are specifically not allowed.  This
> - * is due to an incompatibility with the FS DAX check and
> - * FAULT_FLAG_ALLOW_RETRY.
> - *
> - * In the CMA case: long term pins in a CMA region would unnecessarily fragment
> - * that region.  And so, CMA attempts to migrate the page before pinning, when
> - * FOLL_LONGTERM is specified.
> - *
> - * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
> - * but an additional pin counting system) will be invoked. This is intended for
> - * anything that gets a page reference and then touches page data (for example,
> - * Direct IO). This lets the filesystem know that some non-file-system entity is
> - * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
> - * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
> - * a call to unpin_user_page().
> - *
> - * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
> - * and separate refcounting mechanisms, however, and that means that each has
> - * its own acquire and release mechanisms:
> - *
> - *     FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
> - *
> - *     FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
> - *
> - * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
> - * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
> - * calls applied to them, and that's perfectly OK. This is a constraint on the
> - * callers, not on the pages.)
> - *
> - * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
> - * directly by the caller. That's in order to help avoid mismatches when
> - * releasing pages: get_user_pages*() pages must be released via put_page(),
> - * while pin_user_pages*() pages must be released via unpin_user_page().
> - *
> - * Please see Documentation/core-api/pin_user_pages.rst for more information.
> - */
> -
>   static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
>   {
>   	if (vm_fault & VM_FAULT_OOM)
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index 500e536796ca..0c80a5ad6e6a 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -1003,4 +1003,77 @@ enum fault_flag {
>   
>   typedef unsigned int __bitwise zap_flags_t;
>   
> +/*
> + * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
> + * other. Here is what they mean, and how to use them:
> + *
> + * FOLL_LONGTERM indicates that the page will be held for an indefinite time
> + * period _often_ under userspace control.  This is in contrast to
> + * iov_iter_get_pages(), whose usages are transient.
> + *
> + * FIXME: For pages which are part of a filesystem, mappings are subject to the
> + * lifetime enforced by the filesystem and we need guarantees that longterm
> + * users like RDMA and V4L2 only establish mappings which coordinate usage with
> + * the filesystem.  Ideas for this coordination include revoking the longterm
> + * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
> + * added after the problem with filesystems was found FS DAX VMAs are
> + * specifically failed.  Filesystem pages are still subject to bugs and use of
> + * FOLL_LONGTERM should be avoided on those pages.
> + *
> + * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
> + * Currently only get_user_pages() and get_user_pages_fast() support this flag
> + * and calls to get_user_pages_[un]locked are specifically not allowed.  This
> + * is due to an incompatibility with the FS DAX check and
> + * FAULT_FLAG_ALLOW_RETRY.
> + *
> + * In the CMA case: long term pins in a CMA region would unnecessarily fragment
> + * that region.  And so, CMA attempts to migrate the page before pinning, when
> + * FOLL_LONGTERM is specified.
> + *
> + * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
> + * but an additional pin counting system) will be invoked. This is intended for
> + * anything that gets a page reference and then touches page data (for example,
> + * Direct IO). This lets the filesystem know that some non-file-system entity is
> + * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
> + * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
> + * a call to unpin_user_page().
> + *
> + * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
> + * and separate refcounting mechanisms, however, and that means that each has
> + * its own acquire and release mechanisms:
> + *
> + *     FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
> + *
> + *     FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
> + *
> + * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
> + * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
> + * calls applied to them, and that's perfectly OK. This is a constraint on the
> + * callers, not on the pages.)
> + *
> + * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
> + * directly by the caller. That's in order to help avoid mismatches when
> + * releasing pages: get_user_pages*() pages must be released via put_page(),
> + * while pin_user_pages*() pages must be released via unpin_user_page().
> + *
> + * Please see Documentation/core-api/pin_user_pages.rst for more information.
> + */
> +#define FOLL_WRITE	0x01	/* check pte is writable */
> +#define FOLL_TOUCH	0x02	/* mark page accessed */
> +#define FOLL_GET	0x04	/* do get_page on page */
> +#define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
> +#define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
> +#define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
> +				 * and return without waiting upon it */
> +#define FOLL_NOFAULT	0x80	/* do not fault in pages */
> +#define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
> +#define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
> +#define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
> +#define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
> +#define FOLL_ANON	0x8000	/* don't do file mappings */
> +#define FOLL_LONGTERM	0x10000	/* mapping lifetime is indefinite: see below */
> +#define FOLL_SPLIT_PMD	0x20000	/* split huge pmd before returning */
> +#define FOLL_PIN	0x40000	/* pages must be released via unpin_user_page */
> +#define FOLL_FAST_ONLY	0x80000	/* gup_fast: prevent fall-back to slow gup */
> +
>   #endif /* _LINUX_MM_TYPES_H */
> 
> 

Re: [RFC PATCH 3/4] netfs: Add a function to extract a UBUF or IOVEC into a BVEC iterator

[David Howells]

I updated the commit message to stop using pinning in a general sense:

    netfs: Add a function to extract a UBUF or IOVEC into a BVEC iterator
    
    Add a function to extract the pages from a user-space supplied iterator
    (UBUF- or IOVEC-type) into a BVEC-type iterator, retaining the pages by
    getting a ref on them (WRITE) or pinning them (READ) as we go.
    
    This is useful in three situations:
    
     (1) A userspace thread may have a sibling that unmaps or remaps the
         process's VM during the operation, changing the assignment of the
         pages and potentially causing an error.  Retaining the pages keeps
         some pages around, even if this occurs; futher, we find out at the
         point of extraction if EFAULT is going to be incurred.
    
     (2) Pages might get swapped out/discarded if not retained, so we want to
         retain them to avoid the reload causing a deadlock due to a DIO
         from/to an mmapped region on the same file.
    
     (3) The iterator may get passed to sendmsg() by the filesystem.  If a
         fault occurs, we may get a short write to a TCP stream that's then
         tricky to recover from.
    
    We don't deal with other types of iterator here, leaving it to other
    mechanisms to retain the pages (eg. PG_locked, PG_writeback and the pipe
    lock).

David

Re: [RFC PATCH 4/4] netfs: Add a function to extract an iterator into a scatterlist

[David Howells]

I updated the patch description to mention that pinning may be used instead of
getting a ref:

    netfs: Add a function to extract an iterator into a scatterlist
    
    Provide a function for filling in a scatterlist from the list of pages
    contained in an iterator.
    
    If the iterator is UBUF- or IOBUF-type, the pages have a ref (WRITE) or a
    pin (READ) taken on them.
    
    If the iterator is BVEC-, KVEC- or XARRAY-type, no ref is taken on the
    pages and it is left to the caller to manage their lifetime.  It cannot be
    assumed that a ref can be validly taken, particularly in the case of a KVEC
    iterator.

David

Re: [RFC PATCH 1/4] mm: Move FOLL_* defs to mm_types.h

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [RFC PATCH 2/4] iov_iter: Add a function to extract a page list from an iterator

[Christoph Hellwig]

On Thu, Nov 17, 2022 at 02:54:54PM +0000, David Howells wrote:
> An additional function, iov_iter_extract_mode() is also provided so that the
> mode of retention that will be employed for an iterator can be queried - and
> therefore how the caller should dispose of the pages later.

Any reason to not just add an out paramter to the main function and
return this directly instead of an extra helper?

> +EXPORT_SYMBOL(iov_iter_extract_pages);

get_user_pages_fast, pin_user_pages_fast are very intentionally
EXPORT_SYMBOL_GPL, which should not be bypassed by an iov_* wrapper.

Re: [RFC PATCH 2/4] iov_iter: Add a function to extract a page list from an iterator

[David Howells]

Christoph Hellwig <hch@infradead.org> wrote:

> > +EXPORT_SYMBOL(iov_iter_extract_pages);
> 
> get_user_pages_fast, pin_user_pages_fast are very intentionally
> EXPORT_SYMBOL_GPL, which should not be bypassed by an iov_* wrapper.

Ah, but I'm intending to replace:

	EXPORT_SYMBOL(iov_iter_get_pages2);
	EXPORT_SYMBOL(iov_iter_get_pages_alloc2);

which *aren't* marked _GPL, so you need to argue that one with Al.

David