All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jeffle Xu <jefflexu@linux.alibaba.com>
To: vgoyal@redhat.com, stefanha@redhat.com, miklos@szeredi.hu
Cc: linux-fsdevel@vger.kernel.org, virtio-fs@redhat.com,
	bo.liu@linux.alibaba.com, joseph.qi@linux.alibaba.com
Subject: [virtiofsd PATCH v5 2/2] virtiofsd: support per-file DAX
Date: Thu, 23 Sep 2021 17:35:45 +0800	[thread overview]
Message-ID: <20210923093545.81512-3-jefflexu@linux.alibaba.com> (raw)
In-Reply-To: <20210923093545.81512-1-jefflexu@linux.alibaba.com>

When DAX window is fully utilized and needs to be exapnded to avoid the
performance fluctuation triggered by DAX window recaliming, it may not
be wise to allocate DAX window for files with size smaller than some
specific point, considering from the perspective of reducing memory
overhead.

To maintain one DAX window chunk (e.g., 2MB in size), 32KB
(512 * 64 bytes) memory footprint will be consumed for page descriptors
inside guest. Thus it'd better disable DAX for those files smaller than
32KB, to reduce the demand for DAX window and thus avoid the unworthy
memory overhead.

Thus only flag the file with FUSE_ATTR_DAX when the file size is greater
than 32 KB. The guest will enable DAX only for those files flagged with
FUSE_ATTR_DAX, when virtiofs is mounted with '-o dax=inode'.

To be noted that both FUSE_LOOKUP and FUSE_READDIRPLUS are affected, and
will convey FUSE_ATTR_DAX flag to the guest.

Signed-off-by: Jeffle Xu <jefflexu@linux.alibaba.com>
---
 tools/virtiofsd/passthrough_ll.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
index b76d878509..6893180c6b 100644
--- a/tools/virtiofsd/passthrough_ll.c
+++ b/tools/virtiofsd/passthrough_ll.c
@@ -53,12 +53,26 @@
 #include <sys/syscall.h>
 #include <sys/wait.h>
 #include <sys/xattr.h>
+#include <sys/user.h>
 #include <syslog.h>
 
 #include "qemu/cutils.h"
 #include "passthrough_helpers.h"
 #include "passthrough_seccomp.h"
 
+/*
+ * One page descriptor (64 bytes in size) needs to be maintained for every page
+ * in the DAX window chunk, i.e., there is certain guest memory overhead when
+ * DAX is enabled. Thus disable DAX for those files smaller than this certain
+ * memory overhead if virtiofs is mounted in per-file DAX mode, in which case
+ * the guest page cache will consume less memory when DAX is disabled.
+ */
+#define FUSE_DAX_SHIFT	21
+#define PAGE_DESC_SHIFT	6
+#define FUSE_PERFILE_DAX_SHIFT \
+    (FUSE_DAX_SHIFT - PAGE_SHIFT + PAGE_DESC_SHIFT)
+#define FUSE_PERFILE_DAX_POINT	(1 << FUSE_PERFILE_DAX_SHIFT)
+
 /* Keep track of inode posix locks for each owner. */
 struct lo_inode_plock {
     uint64_t lock_owner;
@@ -1023,6 +1037,11 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
         e->attr_flags |= FUSE_ATTR_SUBMOUNT;
     }
 
+    if (S_ISREG(e->attr.st_mode) &&
+        (e->attr.st_size > FUSE_PERFILE_DAX_POINT)) {
+        e->attr_flags |= FUSE_ATTR_DAX;
+    }
+
     inode = lo_find(lo, &e->attr, mnt_id);
     if (inode) {
         close(newfd);
-- 
2.27.0


WARNING: multiple messages have this Message-ID (diff)
From: Jeffle Xu <jefflexu@linux.alibaba.com>
To: vgoyal@redhat.com, stefanha@redhat.com, miklos@szeredi.hu
Cc: linux-fsdevel@vger.kernel.org, virtio-fs@redhat.com,
	joseph.qi@linux.alibaba.com
Subject: [Virtio-fs] [virtiofsd PATCH v5 2/2] virtiofsd: support per-file DAX
Date: Thu, 23 Sep 2021 17:35:45 +0800	[thread overview]
Message-ID: <20210923093545.81512-3-jefflexu@linux.alibaba.com> (raw)
In-Reply-To: <20210923093545.81512-1-jefflexu@linux.alibaba.com>

When DAX window is fully utilized and needs to be exapnded to avoid the
performance fluctuation triggered by DAX window recaliming, it may not
be wise to allocate DAX window for files with size smaller than some
specific point, considering from the perspective of reducing memory
overhead.

To maintain one DAX window chunk (e.g., 2MB in size), 32KB
(512 * 64 bytes) memory footprint will be consumed for page descriptors
inside guest. Thus it'd better disable DAX for those files smaller than
32KB, to reduce the demand for DAX window and thus avoid the unworthy
memory overhead.

Thus only flag the file with FUSE_ATTR_DAX when the file size is greater
than 32 KB. The guest will enable DAX only for those files flagged with
FUSE_ATTR_DAX, when virtiofs is mounted with '-o dax=inode'.

To be noted that both FUSE_LOOKUP and FUSE_READDIRPLUS are affected, and
will convey FUSE_ATTR_DAX flag to the guest.

Signed-off-by: Jeffle Xu <jefflexu@linux.alibaba.com>
---
 tools/virtiofsd/passthrough_ll.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
index b76d878509..6893180c6b 100644
--- a/tools/virtiofsd/passthrough_ll.c
+++ b/tools/virtiofsd/passthrough_ll.c
@@ -53,12 +53,26 @@
 #include <sys/syscall.h>
 #include <sys/wait.h>
 #include <sys/xattr.h>
+#include <sys/user.h>
 #include <syslog.h>
 
 #include "qemu/cutils.h"
 #include "passthrough_helpers.h"
 #include "passthrough_seccomp.h"
 
+/*
+ * One page descriptor (64 bytes in size) needs to be maintained for every page
+ * in the DAX window chunk, i.e., there is certain guest memory overhead when
+ * DAX is enabled. Thus disable DAX for those files smaller than this certain
+ * memory overhead if virtiofs is mounted in per-file DAX mode, in which case
+ * the guest page cache will consume less memory when DAX is disabled.
+ */
+#define FUSE_DAX_SHIFT	21
+#define PAGE_DESC_SHIFT	6
+#define FUSE_PERFILE_DAX_SHIFT \
+    (FUSE_DAX_SHIFT - PAGE_SHIFT + PAGE_DESC_SHIFT)
+#define FUSE_PERFILE_DAX_POINT	(1 << FUSE_PERFILE_DAX_SHIFT)
+
 /* Keep track of inode posix locks for each owner. */
 struct lo_inode_plock {
     uint64_t lock_owner;
@@ -1023,6 +1037,11 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
         e->attr_flags |= FUSE_ATTR_SUBMOUNT;
     }
 
+    if (S_ISREG(e->attr.st_mode) &&
+        (e->attr.st_size > FUSE_PERFILE_DAX_POINT)) {
+        e->attr_flags |= FUSE_ATTR_DAX;
+    }
+
     inode = lo_find(lo, &e->attr, mnt_id);
     if (inode) {
         close(newfd);
-- 
2.27.0


  parent reply	other threads:[~2021-09-23  9:35 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-23  9:25 [PATCH v5 0/5] fuse,virtiofs: support per-file DAX Jeffle Xu
2021-09-23  9:25 ` [Virtio-fs] " Jeffle Xu
2021-09-23  9:25 ` [PATCH v5 1/5] fuse: add fuse_should_enable_dax() helper Jeffle Xu
2021-09-23  9:25   ` [Virtio-fs] " Jeffle Xu
2021-09-23 18:58   ` Vivek Goyal
2021-09-23 18:58     ` [Virtio-fs] " Vivek Goyal
2021-09-23  9:25 ` [PATCH v5 2/5] fuse: make DAX mount option a tri-state Jeffle Xu
2021-09-23  9:25   ` [Virtio-fs] " Jeffle Xu
2021-09-23 19:02   ` Vivek Goyal
2021-09-23 19:02     ` [Virtio-fs] " Vivek Goyal
2021-09-23 22:26     ` Dave Chinner
2021-09-23 22:26       ` [Virtio-fs] " Dave Chinner
2021-09-24  1:02       ` Vivek Goyal
2021-09-24  1:02         ` [Virtio-fs] " Vivek Goyal
2021-09-24  3:06         ` JeffleXu
2021-09-24  3:06           ` [Virtio-fs] " JeffleXu
2021-09-24 12:43           ` Vivek Goyal
2021-09-24 12:43             ` [Virtio-fs] " Vivek Goyal
2021-09-26  2:06             ` JeffleXu
2021-09-26  2:06               ` [Virtio-fs] " JeffleXu
2021-09-27  0:21         ` Dave Chinner
2021-09-27  0:21           ` [Virtio-fs] " Dave Chinner
2021-09-27  2:28           ` JeffleXu
2021-09-27  2:28             ` [Virtio-fs] " JeffleXu
2021-09-28  3:44             ` Dave Chinner
2021-09-28  3:44               ` [Virtio-fs] " Dave Chinner
2021-09-28  5:17               ` JeffleXu
2021-09-28  5:17                 ` [Virtio-fs] " JeffleXu
2021-09-28 14:34                 ` Vivek Goyal
2021-09-28 14:34                   ` [Virtio-fs] " Vivek Goyal
2021-09-27 15:32           ` Vivek Goyal
2021-09-27 15:32             ` [Virtio-fs] " Vivek Goyal
2021-09-28  3:23             ` Dave Chinner
2021-09-28  3:23               ` [Virtio-fs] " Dave Chinner
2021-09-23  9:25 ` [PATCH v5 3/5] fuse: support per-file DAX Jeffle Xu
2021-09-23  9:25   ` [Virtio-fs] " Jeffle Xu
2021-09-23  9:25 ` [PATCH v5 4/5] fuse: enable " Jeffle Xu
2021-09-23  9:25   ` [Virtio-fs] " Jeffle Xu
2021-09-23  9:25 ` [PATCH v5 5/5] fuse: mark inode DONT_CACHE when per-file DAX hint changes Jeffle Xu
2021-09-23  9:25   ` [Virtio-fs] " Jeffle Xu
2021-09-23  9:35 ` [virtiofsd PATCH v5 0/2] virtiofsd: support per-file DAX Jeffle Xu
2021-09-23  9:35   ` [Virtio-fs] " Jeffle Xu
2021-09-23  9:35   ` [virtiofsd PATCH v5 1/2] virtiofsd: add FUSE_ATTR_DAX to fuse protocol Jeffle Xu
2021-09-23  9:35     ` [Virtio-fs] " Jeffle Xu
2021-09-23  9:35   ` Jeffle Xu [this message]
2021-09-23  9:35     ` [Virtio-fs] [virtiofsd PATCH v5 2/2] virtiofsd: support per-file DAX Jeffle Xu
2021-09-23 18:57 ` [PATCH v5 0/5] fuse,virtiofs: " Vivek Goyal
2021-09-23 18:57   ` [Virtio-fs] " Vivek Goyal
2021-10-27  3:42   ` JeffleXu
2021-10-27  3:42     ` [Virtio-fs] " JeffleXu
2021-10-27 14:45     ` Vivek Goyal
2021-10-27 14:45       ` [Virtio-fs] " Vivek Goyal

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=20210923093545.81512-3-jefflexu@linux.alibaba.com \
    --to=jefflexu@linux.alibaba.com \
    --cc=bo.liu@linux.alibaba.com \
    --cc=joseph.qi@linux.alibaba.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=stefanha@redhat.com \
    --cc=vgoyal@redhat.com \
    --cc=virtio-fs@redhat.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.