pve-qemu/debian/patches/pve/0018-PVE-add-optional-buffer-size-to-QEMUFile.patch

From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Wolfgang Bumiller <w.bumiller@proxmox.com>
Date: Mon, 4 May 2020 11:05:08 +0200
Subject: [PATCH] PVE: add optional buffer size to QEMUFile

So we can use a 4M buffer for savevm-async which should
increase performance storing the state onto ceph.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
[increase max IOV count in QEMUFile to actually write more data]
Signed-off-by: Stefan Reiter <s.reiter@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
[FE: adapt to removal of QEMUFileOps]
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
---
 migration/qemu-file.c    | 50 +++++++++++++++++++++++++++-------------
 migration/qemu-file.h    |  2 ++
 migration/savevm-async.c |  5 ++--
 3 files changed, 39 insertions(+), 18 deletions(-)

diff --git a/migration/qemu-file.c b/migration/qemu-file.c
index a10882d47f..19c1de0472 100644
--- a/migration/qemu-file.c
+++ b/migration/qemu-file.c
@@ -35,8 +35,8 @@
 #include "rdma.h"
 #include "io/channel-file.h"
 
-#define IO_BUF_SIZE 32768
-#define MAX_IOV_SIZE MIN_CONST(IOV_MAX, 64)
+#define DEFAULT_IO_BUF_SIZE 32768
+#define MAX_IOV_SIZE MIN_CONST(IOV_MAX, 256)
 
 struct QEMUFile {
     QIOChannel *ioc;
@@ -44,7 +44,8 @@ struct QEMUFile {
 
     int buf_index;
     int buf_size; /* 0 when writing */
-    uint8_t buf[IO_BUF_SIZE];
+    size_t buf_allocated_size;
+    uint8_t *buf;
 
     DECLARE_BITMAP(may_free, MAX_IOV_SIZE);
     struct iovec iov[MAX_IOV_SIZE];
@@ -101,7 +102,9 @@ int qemu_file_shutdown(QEMUFile *f)
     return 0;
 }
 
-static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
+static QEMUFile *qemu_file_new_impl(QIOChannel *ioc,
+                                    bool is_writable,
+                                    size_t buffer_size)
 {
     QEMUFile *f;
 
@@ -110,6 +113,8 @@ static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
     object_ref(ioc);
     f->ioc = ioc;
     f->is_writable = is_writable;
+    f->buf_allocated_size = buffer_size;
+    f->buf = malloc(buffer_size);
 
     return f;
 }
@@ -120,17 +125,27 @@ static QEMUFile *qemu_file_new_impl(QIOChannel *ioc, bool is_writable)
  */
 QEMUFile *qemu_file_get_return_path(QEMUFile *f)
 {
-    return qemu_file_new_impl(f->ioc, !f->is_writable);
+    return qemu_file_new_impl(f->ioc, !f->is_writable, DEFAULT_IO_BUF_SIZE);
 }
 
 QEMUFile *qemu_file_new_output(QIOChannel *ioc)
 {
-    return qemu_file_new_impl(ioc, true);
+    return qemu_file_new_impl(ioc, true, DEFAULT_IO_BUF_SIZE);
+}
+
+QEMUFile *qemu_file_new_output_sized(QIOChannel *ioc, size_t buffer_size)
+{
+    return qemu_file_new_impl(ioc, true, buffer_size);
 }
 
 QEMUFile *qemu_file_new_input(QIOChannel *ioc)
 {
-    return qemu_file_new_impl(ioc, false);
+    return qemu_file_new_impl(ioc, false, DEFAULT_IO_BUF_SIZE);
+}
+
+QEMUFile *qemu_file_new_input_sized(QIOChannel *ioc, size_t buffer_size)
+{
+    return qemu_file_new_impl(ioc, false, buffer_size);
 }
 
 /*
@@ -328,7 +343,7 @@ static ssize_t coroutine_mixed_fn qemu_fill_buffer(QEMUFile *f)
     do {
         len = qio_channel_read(f->ioc,
                                (char *)f->buf + pending,
-                               IO_BUF_SIZE - pending,
+                               f->buf_allocated_size - pending,
                                &local_error);
         if (len == QIO_CHANNEL_ERR_BLOCK) {
             if (qemu_in_coroutine()) {
@@ -368,6 +383,9 @@ int qemu_fclose(QEMUFile *f)
         ret = ret2;
     }
     g_clear_pointer(&f->ioc, object_unref);
+
+    free(f->buf);
+
     error_free(f->last_error_obj);
     g_free(f);
     trace_qemu_file_fclose();
@@ -416,7 +434,7 @@ static void add_buf_to_iovec(QEMUFile *f, size_t len)
 {
     if (!add_to_iovec(f, f->buf + f->buf_index, len, false)) {
         f->buf_index += len;
-        if (f->buf_index == IO_BUF_SIZE) {
+        if (f->buf_index == f->buf_allocated_size) {
             qemu_fflush(f);
         }
     }
@@ -441,7 +459,7 @@ void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
     }
 
     while (size > 0) {
-        l = IO_BUF_SIZE - f->buf_index;
+        l = f->buf_allocated_size - f->buf_index;
         if (l > size) {
             l = size;
         }
@@ -587,8 +605,8 @@ size_t coroutine_mixed_fn qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t si
     size_t index;
 
     assert(!qemu_file_is_writable(f));
-    assert(offset < IO_BUF_SIZE);
-    assert(size <= IO_BUF_SIZE - offset);
+    assert(offset < f->buf_allocated_size);
+    assert(size <= f->buf_allocated_size - offset);
 
     /* The 1st byte to read from */
     index = f->buf_index + offset;
@@ -638,7 +656,7 @@ size_t coroutine_mixed_fn qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size
         size_t res;
         uint8_t *src;
 
-        res = qemu_peek_buffer(f, &src, MIN(pending, IO_BUF_SIZE), 0);
+        res = qemu_peek_buffer(f, &src, MIN(pending, f->buf_allocated_size), 0);
         if (res == 0) {
             return done;
         }
@@ -672,7 +690,7 @@ size_t coroutine_mixed_fn qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size
  */
 size_t coroutine_mixed_fn qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
 {
-    if (size < IO_BUF_SIZE) {
+    if (size < f->buf_allocated_size) {
         size_t res;
         uint8_t *src = NULL;
 
@@ -697,7 +715,7 @@ int coroutine_mixed_fn qemu_peek_byte(QEMUFile *f, int offset)
     int index = f->buf_index + offset;
 
     assert(!qemu_file_is_writable(f));
-    assert(offset < IO_BUF_SIZE);
+    assert(offset < f->buf_allocated_size);
 
     if (index >= f->buf_size) {
         qemu_fill_buffer(f);
@@ -811,7 +829,7 @@ static int qemu_compress_data(z_stream *stream, uint8_t *dest, size_t dest_len,
 ssize_t qemu_put_compression_data(QEMUFile *f, z_stream *stream,
                                   const uint8_t *p, size_t size)
 {
-    ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
+    ssize_t blen = f->buf_allocated_size - f->buf_index - sizeof(int32_t);
 
     if (blen < compressBound(size)) {
         return -1;
diff --git a/migration/qemu-file.h b/migration/qemu-file.h
index 32fd4a34fd..36a0cd8cc8 100644
--- a/migration/qemu-file.h
+++ b/migration/qemu-file.h
@@ -30,7 +30,9 @@
 #include "io/channel.h"
 
 QEMUFile *qemu_file_new_input(QIOChannel *ioc);
+QEMUFile *qemu_file_new_input_sized(QIOChannel *ioc, size_t buffer_size);
 QEMUFile *qemu_file_new_output(QIOChannel *ioc);
+QEMUFile *qemu_file_new_output_sized(QIOChannel *ioc, size_t buffer_size);
 int qemu_fclose(QEMUFile *f);
 
 /*
diff --git a/migration/savevm-async.c b/migration/savevm-async.c
index 1af32604c7..be2035cd2e 100644
--- a/migration/savevm-async.c
+++ b/migration/savevm-async.c
@@ -386,7 +386,7 @@ void qmp_savevm_start(const char *statefile, Error **errp)
 
     QIOChannel *ioc = QIO_CHANNEL(qio_channel_savevm_async_new(snap_state.target,
                                                                &snap_state.bs_pos));
-    snap_state.file = qemu_file_new_output(ioc);
+    snap_state.file = qemu_file_new_output_sized(ioc, 4 * 1024 * 1024);
 
     if (!snap_state.file) {
         error_set(errp, ERROR_CLASS_GENERIC_ERROR, "failed to open '%s'", statefile);
@@ -510,7 +510,8 @@ int load_snapshot_from_blockdev(const char *filename, Error **errp)
     blk_op_block_all(be, blocker);
 
     /* restore the VM state */
-    f = qemu_file_new_input(QIO_CHANNEL(qio_channel_savevm_async_new(be, &bs_pos)));
+    f = qemu_file_new_input_sized(QIO_CHANNEL(qio_channel_savevm_async_new(be, &bs_pos)),
+                                  4 * 1024 * 1024);
     if (!f) {
         error_setg(errp, "Could not open VM state file");
         goto the_end;