From: wangtao <tao.wangtao@honor.com>
To: <sumit.semwal@linaro.org>, <christian.koenig@amd.com>,
<kraxel@redhat.com>, <vivek.kasireddy@intel.com>,
<viro@zeniv.linux.org.uk>, <brauner@kernel.org>,
<hughd@google.com>, <akpm@linux-foundation.org>,
<amir73il@gmail.com>
Cc: <benjamin.gaignard@collabora.com>, <Brian.Starkey@arm.com>,
<jstultz@google.com>, <tjmercier@google.com>, <jack@suse.cz>,
<baolin.wang@linux.alibaba.com>, <linux-media@vger.kernel.org>,
<dri-devel@lists.freedesktop.org>,
<linaro-mm-sig@lists.linaro.org>, <linux-kernel@vger.kernel.org>,
<linux-fsdevel@vger.kernel.org>, <linux-mm@kvack.org>,
<bintian.wang@honor.com>, <yipengxiang@honor.com>,
<liulu.liu@honor.com>, <feng.han@honor.com>,
wangtao <tao.wangtao@honor.com>
Subject: [PATCH v3 1/4] fs: allow cross-FS copy_file_range for memory-backed files
Date: Fri, 30 May 2025 18:39:38 +0800 [thread overview]
Message-ID: <20250530103941.11092-2-tao.wangtao@honor.com> (raw)
In-Reply-To: <20250530103941.11092-1-tao.wangtao@honor.com>
Memory-backed files can optimize copy performance via
copy_file_range callbacks. Compared to mmap&read: reduces
GUP (get_user_pages) overhead; vs sendfile/splice: eliminates
one memory copy; supports dmabuf zero-copy implementation.
Signed-off-by: wangtao <tao.wangtao@honor.com>
---
fs/read_write.c | 71 +++++++++++++++++++++++++++++++++-------------
include/linux/fs.h | 2 ++
2 files changed, 54 insertions(+), 19 deletions(-)
diff --git a/fs/read_write.c b/fs/read_write.c
index bb0ed26a0b3a..591c6db7b785 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1469,6 +1469,20 @@ COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
}
#endif
+static inline bool is_copy_memory_file_to_file(struct file *file_in,
+ struct file *file_out)
+{
+ return (file_in->f_op->fop_flags & FOP_MEMORY_FILE) &&
+ file_in->f_op->copy_file_range && file_out->f_op->write_iter;
+}
+
+static inline bool is_copy_file_to_memory_file(struct file *file_in,
+ struct file *file_out)
+{
+ return (file_out->f_op->fop_flags & FOP_MEMORY_FILE) &&
+ file_in->f_op->read_iter && file_out->f_op->copy_file_range;
+}
+
/*
* Performs necessary checks before doing a file copy
*
@@ -1484,11 +1498,23 @@ static int generic_copy_file_checks(struct file *file_in, loff_t pos_in,
struct inode *inode_out = file_inode(file_out);
uint64_t count = *req_count;
loff_t size_in;
+ bool splice = flags & COPY_FILE_SPLICE;
+ bool has_memory_file;
int ret;
- ret = generic_file_rw_checks(file_in, file_out);
- if (ret)
- return ret;
+ /* Skip generic checks, allow cross-sb copies for dma-buf/tmpfs */
+ has_memory_file = is_copy_memory_file_to_file(file_in, file_out) ||
+ is_copy_file_to_memory_file(file_in, file_out);
+ if (!splice && has_memory_file) {
+ if (!(file_in->f_mode & FMODE_READ) ||
+ !(file_out->f_mode & FMODE_WRITE) ||
+ (file_out->f_flags & O_APPEND))
+ return -EBADF;
+ } else {
+ ret = generic_file_rw_checks(file_in, file_out);
+ if (ret)
+ return ret;
+ }
/*
* We allow some filesystems to handle cross sb copy, but passing
@@ -1500,7 +1526,7 @@ static int generic_copy_file_checks(struct file *file_in, loff_t pos_in,
* and several different sets of file_operations, but they all end up
* using the same ->copy_file_range() function pointer.
*/
- if (flags & COPY_FILE_SPLICE) {
+ if (splice || has_memory_file) {
/* cross sb splice is allowed */
} else if (file_out->f_op->copy_file_range) {
if (file_in->f_op->copy_file_range !=
@@ -1581,23 +1607,30 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
* same sb using clone, but for filesystems where both clone and copy
* are supported (e.g. nfs,cifs), we only call the copy method.
*/
- if (!splice && file_out->f_op->copy_file_range) {
- ret = file_out->f_op->copy_file_range(file_in, pos_in,
- file_out, pos_out,
- len, flags);
- } else if (!splice && file_in->f_op->remap_file_range && samesb) {
- ret = file_in->f_op->remap_file_range(file_in, pos_in,
- file_out, pos_out,
- min_t(loff_t, MAX_RW_COUNT, len),
- REMAP_FILE_CAN_SHORTEN);
- /* fallback to splice */
- if (ret <= 0)
+ if (!splice) {
+ if (is_copy_memory_file_to_file(file_in, file_out)) {
+ ret = file_in->f_op->copy_file_range(file_in, pos_in,
+ file_out, pos_out, len, flags);
+ } else if (is_copy_file_to_memory_file(file_in, file_out)) {
+ ret = file_out->f_op->copy_file_range(file_in, pos_in,
+ file_out, pos_out, len, flags);
+ } else if (file_out->f_op->copy_file_range) {
+ ret = file_out->f_op->copy_file_range(file_in, pos_in,
+ file_out, pos_out,
+ len, flags);
+ } else if (file_in->f_op->remap_file_range && samesb) {
+ ret = file_in->f_op->remap_file_range(file_in, pos_in,
+ file_out, pos_out,
+ min_t(loff_t, MAX_RW_COUNT, len),
+ REMAP_FILE_CAN_SHORTEN);
+ /* fallback to splice */
+ if (ret <= 0)
+ splice = true;
+ } else if (samesb) {
+ /* Fallback to splice for same sb copy for backward compat */
splice = true;
- } else if (samesb) {
- /* Fallback to splice for same sb copy for backward compat */
- splice = true;
+ }
}
-
file_end_write(file_out);
if (!splice)
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 016b0fe1536e..37df1b497418 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2187,6 +2187,8 @@ struct file_operations {
#define FOP_ASYNC_LOCK ((__force fop_flags_t)(1 << 6))
/* File system supports uncached read/write buffered IO */
#define FOP_DONTCACHE ((__force fop_flags_t)(1 << 7))
+/* Supports cross-FS copy_file_range for memory file */
+#define FOP_MEMORY_FILE ((__force fop_flags_t)(1 << 8))
/* Wrap a directory iterator that needs exclusive inode access */
int wrap_directory_iterator(struct file *, struct dir_context *,
--
2.17.1
next prev parent reply other threads:[~2025-05-30 10:41 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-05-30 10:39 [PATCH v3 0/4] Optimizing disk file & dmabuf copies via copy_file_range wangtao
2025-05-30 10:39 ` wangtao [this message]
2025-05-30 13:43 ` [PATCH v3 1/4] fs: allow cross-FS copy_file_range for memory-backed files Amir Goldstein
2025-06-03 1:28 ` wangtao
2025-05-30 10:39 ` [PATCH v3 2/4] dmabuf: Implement copy_file_range for dmabuf wangtao
2025-05-30 10:39 ` [PATCH v3 3/4] udmabuf: Implement udmabuf rw_file callback wangtao
2025-05-30 14:24 ` kernel test robot
2025-06-03 1:32 ` wangtao
2025-05-30 10:39 ` [PATCH v3 4/4] dmabuf:system_heap Implement system_heap exporter's " wangtao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250530103941.11092-2-tao.wangtao@honor.com \
--to=tao.wangtao@honor.com \
--cc=Brian.Starkey@arm.com \
--cc=akpm@linux-foundation.org \
--cc=amir73il@gmail.com \
--cc=baolin.wang@linux.alibaba.com \
--cc=benjamin.gaignard@collabora.com \
--cc=bintian.wang@honor.com \
--cc=brauner@kernel.org \
--cc=christian.koenig@amd.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=feng.han@honor.com \
--cc=hughd@google.com \
--cc=jack@suse.cz \
--cc=jstultz@google.com \
--cc=kraxel@redhat.com \
--cc=linaro-mm-sig@lists.linaro.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-media@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=liulu.liu@honor.com \
--cc=sumit.semwal@linaro.org \
--cc=tjmercier@google.com \
--cc=viro@zeniv.linux.org.uk \
--cc=vivek.kasireddy@intel.com \
--cc=yipengxiang@honor.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox