* [PATCH] btrfs-progs: send: sync splice buf size with kernel when proto 2
@ 2022-10-17 3:52 Wang Yugui
2022-10-18 13:48 ` David Sterba
0 siblings, 1 reply; 2+ messages in thread
From: Wang Yugui @ 2022-10-17 3:52 UTC (permalink / raw)
To: linux-btrfs; +Cc: Wang Yugui
When 'btrfs send --proto 2', the max buffer in kernel is changed from
BTRFS_SEND_BUF_SIZE_V1(SZ_64K) to (SZ_16K + BTRFS_MAX_COMPRESSED).
The performance is improved when we use the same buf size in btrfs-progs.
without this patch: 57.96s
with this patch: 48.44s
bigger buf size(SZ_512K) is tested too. but it help 'btrfs send --proto 2' /
'btrfs send --proto 1' just little. so we will not use bigger buffer.
Signed-off-by: Wang Yugui <wangyugui@e16-tech.com>
---
cmds/send.c | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/cmds/send.c b/cmds/send.c
index ec61aaf4..d18a31a1 100644
--- a/cmds/send.c
+++ b/cmds/send.c
@@ -37,7 +37,10 @@
#include "cmds/commands.h"
#include "ioctl.h"
-#define SEND_BUFFER_SIZE SZ_64K
+#define BTRFS_SEND_BUF_SIZE_V1 (SZ_64K)
+#define BTRFS_MAX_COMPRESSED (SZ_128K)
+#define BTRFS_SEND_BUF_SIZE_V2 (SZ_16K + BTRFS_MAX_COMPRESSED)
+
struct btrfs_send {
int send_fd;
@@ -201,11 +204,18 @@ static void *read_sent_data(void *arg)
struct btrfs_send *sctx = (struct btrfs_send*)arg;
while (1) {
+ size_t splice_buf_size = BTRFS_SEND_BUF_SIZE_V1;
ssize_t sbytes;
+ if(sctx->proto > 1){
+ splice_buf_size = BTRFS_SEND_BUF_SIZE_V2;
+
+ /* try to change pipe buffer size */
+ fcntl(sctx->dump_fd, F_SETPIPE_SZ, splice_buf_size);
+ }
/* Source is a pipe, output is either file or stdout */
sbytes = splice(sctx->send_fd, NULL, sctx->dump_fd,
- NULL, SEND_BUFFER_SIZE, SPLICE_F_MORE);
+ NULL, splice_buf_size, SPLICE_F_MORE);
if (sbytes < 0) {
ret = -errno;
error("failed to read stream from kernel: %m");
@@ -261,6 +271,9 @@ static int do_send(struct btrfs_send *send, u64 parent_root_id,
*/
io_send.version = send->proto;
io_send.flags |= BTRFS_SEND_FLAG_VERSION;
+
+ fcntl(pipefd[0], F_SETPIPE_SZ, BTRFS_SEND_BUF_SIZE_V2);
+ fcntl(pipefd[1], F_SETPIPE_SZ, BTRFS_SEND_BUF_SIZE_V2);
}
if (!ret)
--
2.36.2
^ permalink raw reply related [flat|nested] 2+ messages in thread* Re: [PATCH] btrfs-progs: send: sync splice buf size with kernel when proto 2
2022-10-17 3:52 [PATCH] btrfs-progs: send: sync splice buf size with kernel when proto 2 Wang Yugui
@ 2022-10-18 13:48 ` David Sterba
0 siblings, 0 replies; 2+ messages in thread
From: David Sterba @ 2022-10-18 13:48 UTC (permalink / raw)
To: Wang Yugui; +Cc: linux-btrfs
On Mon, Oct 17, 2022 at 11:52:31AM +0800, Wang Yugui wrote:
> When 'btrfs send --proto 2', the max buffer in kernel is changed from
> BTRFS_SEND_BUF_SIZE_V1(SZ_64K) to (SZ_16K + BTRFS_MAX_COMPRESSED).
>
> The performance is improved when we use the same buf size in btrfs-progs.
> without this patch: 57.96s
> with this patch: 48.44s
>
> bigger buf size(SZ_512K) is tested too. but it help 'btrfs send --proto 2' /
> 'btrfs send --proto 1' just little. so we will not use bigger buffer.
>
> Signed-off-by: Wang Yugui <wangyugui@e16-tech.com>
Added to devel, thanks.
> cmds/send.c | 17 +++++++++++++++--
> 1 file changed, 15 insertions(+), 2 deletions(-)
>
> diff --git a/cmds/send.c b/cmds/send.c
> index ec61aaf4..d18a31a1 100644
> --- a/cmds/send.c
> +++ b/cmds/send.c
> @@ -37,7 +37,10 @@
> #include "cmds/commands.h"
> #include "ioctl.h"
>
> -#define SEND_BUFFER_SIZE SZ_64K
> +#define BTRFS_SEND_BUF_SIZE_V1 (SZ_64K)
> +#define BTRFS_MAX_COMPRESSED (SZ_128K)
> +#define BTRFS_SEND_BUF_SIZE_V2 (SZ_16K + BTRFS_MAX_COMPRESSED)
Such define is missing from kernel, right? It's hidden in
btrfs_ioctl_send as
sctx->send_max_size = ALIGN(SZ_16K + BTRFS_MAX_COMPRESSED, PAGE_SIZE);
would be nice to have that as a separate define.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2022-10-18 13:48 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2022-10-17 3:52 [PATCH] btrfs-progs: send: sync splice buf size with kernel when proto 2 Wang Yugui
2022-10-18 13:48 ` David Sterba
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).