[Spice-devel] [PATCH 5/7] server/tests/replay: introduce
Jonathon Jongsma
jjongsma at redhat.com
Fri Aug 14 09:43:45 PDT 2015
On Thu, 2015-08-13 at 16:25 +0100, Frediano Ziglio wrote:
> From: Alon Levy <alon at pobox.com>
>
> usage: replay <cmdfile> <port> <client command line>
This usage doesn't really match the implementation below. It should be
something more like:
replay -p <port> -c <client command line> <cmdfile>
While we're on the subject, though, can we use a more specific name?
e.g. spice-replay
>
> will run the commands from cmdfile ignoring timestamps, right after a
> connection is established from the client, and will SIGINT the client
> on end of cmdfile, and exit itself after waiting for the client.
>
> spicy-stats from spice-gtk is useful for testing, it prints the summary
> of the traffic on each channel.
>
> You can also run with no client by doing:
> replay <cmdfile>
>
> For example, the 300 MB file (compressed to 4 MB with xz -9) available
> at [1] produces the following output:
>
> spicy-stats total bytes read:
> total bytes read:
> inputs: 214
> display: 1968983
> cursor: 390
> main: 256373
>
> You could run it directly like so:
> curl http://annarchy.freedesktop.org/~alon/win7_boot_shutdown.cmd.xz | xzcat | server/tests/replay - 12345 `which spicy-stats` -h localhost -p 12345
Again, command above doesn't seem to match the implementation.
>
> Known Problems:
> * Implementation is wrong. Should do a single device->host conversion
> (i.e. get_virt), and then marshall/demarshall that (i.e. RedDrawable).
> * segfault on file read done resulting in the above spicy-stats not
> being reproducable (well, up to 1% yes).
>
> [1] http://annarchy.freedesktop.org/~alon/win7_boot_shutdown.cmd.xz
>
> Now based on glib including using an asyncqueue for reading the playback
> file, and proper freeing of the allocated commands, with --slow,
> --compression and a progress timer, and doesn't use more then nsurfaces.3
This last paragraph seems to be addressing differences to a previous
version of the patch. We can probably incorporate this into the above
commit log or just drop it.
> ---
> server/red_replay_qxl.c | 7 +-
> server/red_replay_qxl.h | 17 +++
> server/tests/Makefile.am | 9 ++
> server/tests/replay.c | 346 +++++++++++++++++++++++++++++++++++++++++++++++
> 4 files changed, 377 insertions(+), 2 deletions(-)
> create mode 100644 server/tests/replay.c
>
> diff --git a/server/red_replay_qxl.c b/server/red_replay_qxl.c
> index 663c6ed..e52ee9c 100644
> --- a/server/red_replay_qxl.c
> +++ b/server/red_replay_qxl.c
> @@ -186,13 +186,15 @@ static replay_t read_binary(SpiceReplay *replay, const char *prefix, size_t *siz
> **buf, size_t base_size)
> {
> char template[1024];
> - int with_zlib;
> + int with_zlib = -1;
> int zlib_size;
> uint8_t *zlib_buffer;
> z_stream strm;
>
> snprintf(template, sizeof(template), "binary %%d %s %%ld:", prefix);
> - replay_fscanf(replay, template, &with_zlib, size);
> + if (replay_fscanf(replay, template, &with_zlib, size) == REPLAY_EOF)
> + return REPLAY_EOF;
> +
> if (*buf == NULL) {
> *buf = malloc(*size + base_size);
> if (*buf == NULL) {
> @@ -207,6 +209,7 @@ static replay_t read_binary(SpiceReplay *replay, const char *prefix, size_t *siz
> hexdump(*buf + base_size, *size);
> }
> #else
> + spice_return_val_if_fail(with_zlib != -1, REPLAY_EOF);
> if (with_zlib) {
> int ret;
>
> diff --git a/server/red_replay_qxl.h b/server/red_replay_qxl.h
> index 77a37e0..a89b3d4 100644
> --- a/server/red_replay_qxl.h
> +++ b/server/red_replay_qxl.h
> @@ -1,3 +1,20 @@
> +/* -*- Mode: C; c-basic-offset: 4; indent-tabs-mode: nil -*- */
> +/*
> + Copyright (C) 2009-2015 Red Hat, Inc.
> +
> + This library is free software; you can redistribute it and/or
> + modify it under the terms of the GNU Lesser General Public
> + License as published by the Free Software Foundation; either
> + version 2.1 of the License, or (at your option) any later version.
> +
> + This library is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + Lesser General Public License for more details.
> +
> + You should have received a copy of the GNU Lesser General Public
> + License along with this library; if not, see <http://www.gnu.org/licenses/>.
> +*/
> #ifndef RED_REPLAY_QXL_H
> #define RED_REPLAY_QXL_H
>
> diff --git a/server/tests/Makefile.am b/server/tests/Makefile.am
> index 32c97ce..4fa03fe 100644
> --- a/server/tests/Makefile.am
> +++ b/server/tests/Makefile.am
> @@ -40,6 +40,7 @@ noinst_PROGRAMS = \
> test_two_servers \
> test_vdagent \
> test_display_width_stride \
> + replay \
> $(NULL)
>
> test_vdagent_SOURCES = \
> @@ -101,3 +102,11 @@ test_display_width_stride_SOURCES = \
> test_display_base.h \
> test_display_width_stride.c \
> $(NULL)
> +
> +replay_SOURCES = \
> + replay.c \
> + test_display_base.h \
> + basic_event_loop.c \
> + basic_event_loop.h \
> + test_util.h \
> + $(NULL)
> diff --git a/server/tests/replay.c b/server/tests/replay.c
> new file mode 100644
> index 0000000..f1f1098
> --- /dev/null
> +++ b/server/tests/replay.c
> @@ -0,0 +1,346 @@
> +/* Replay a previously recorded file (via SPICE_WORKER_RECORD_FILENAME)
> + */
> +
> +#include <string.h>
> +#include <stdlib.h>
> +#include <stdio.h>
> +#include <strings.h>
> +#include <sys/types.h>
> +#include <signal.h>
> +#include <unistd.h>
> +#include <pthread.h>
> +#include <sys/wait.h>
> +#include <fcntl.h>
> +#include <glib.h>
> +
> +#include <spice/macros.h>
> +#include "red_replay_qxl.h"
> +#include "test_display_base.h"
> +#include "common/log.h"
> +
> +static SpiceCoreInterface *core;
> +static SpiceServer *server;
> +static SpiceReplay *replay;
> +static QXLWorker *qxl_worker = NULL;
> +static gboolean started = FALSE;
> +static QXLInstance display_sin = { 0, };
> +static gint slow = 0;
> +static pid_t client_pid;
> +static GMainLoop *loop = NULL;
> +static GAsyncQueue *aqueue = NULL;
> +static long total_size;
> +
> +static GMutex mutex;
> +static guint fill_source_id = 0;
> +
> +
> +#define MEM_SLOT_GROUP_ID 0
> +
> +/* Parts cribbed from spice-display.h/.c/qxl.c */
> +
> +static QXLDevMemSlot slot = {
> +.slot_group_id = MEM_SLOT_GROUP_ID,
> +.slot_id = 0,
> +.generation = 0,
> +.virt_start = 0,
> +.virt_end = ~0,
> +.addr_delta = 0,
> +.qxl_ram_size = ~0,
> +};
> +
> +static void attach_worker(QXLInstance *qin, QXLWorker *_qxl_worker)
> +{
> + static int count = 0;
> + if (++count > 1) {
> + printf("%s ignored\n", __func__);
g_warning()?
> + return;
> + }
> + printf("%s\n", __func__);
Is this necessary? perhaps use g_debug() so it doesn't print by default.
> + qxl_worker = _qxl_worker;
> + spice_qxl_add_memslot(qin, &slot);
> + spice_server_vm_start(server);
> +}
> +
> +static void set_compression_level(QXLInstance *qin, int level)
> +{
> + printf("%s\n", __func__);
g_debug()?
> +}
> +
> +static void set_mm_time(QXLInstance *qin, uint32_t mm_time)
> +{
> +}
> +
> +// same as qemu/ui/spice-display.h
> +#define MAX_SURFACE_NUM 1024
> +
> +static void get_init_info(QXLInstance *qin, QXLDevInitInfo *info)
> +{
> + bzero(info, sizeof(*info));
> + info->num_memslots = 1;
> + info->num_memslots_groups = 1;
> + info->memslot_id_bits = 1;
> + info->memslot_gen_bits = 1;
> + info->n_surfaces = MAX_SURFACE_NUM;
> +}
> +
> +static gboolean fill_queue_idle(gpointer user_data)
> +{
> + gboolean keep = FALSE;
> +
> + while (g_async_queue_length(aqueue) < 50) {
> + QXLCommandExt *cmd = spice_replay_next_cmd(replay, qxl_worker);
> + if (!cmd) {
> + g_async_queue_push(aqueue, GINT_TO_POINTER(-1));
> + goto end;
> + }
> +
> + if (slow)
> + g_usleep(slow);
> +
> + g_async_queue_push(aqueue, cmd);
> + }
> +
> +end:
> + if (!keep) {
> + g_mutex_lock(&mutex);
> + fill_source_id = 0;
> + g_mutex_unlock(&mutex);
> + }
> + spice_qxl_wakeup(&display_sin);
> +
> + return keep;
> +}
> +
> +static void fill_queue(void)
> +{
> + g_mutex_lock(&mutex);
> +
> + if (!started)
> + goto end;
> +
> + if (fill_source_id != 0)
> + goto end;
> +
> + fill_source_id = g_idle_add(fill_queue_idle, NULL);
> +
> +end:
> + g_mutex_unlock(&mutex);
> +}
> +
> +
> +// called from spice_server thread (i.e. red_worker thread)
> +static int get_command(QXLInstance *qin, QXLCommandExt *ext)
> +{
> + QXLCommandExt *cmd;
> +
> + if (g_async_queue_length(aqueue) == 0) {
> + /* could use a gcondition ? */
> + fill_queue();
> + return FALSE;
> + }
> +
> + cmd = g_async_queue_try_pop(aqueue);
> + if (GPOINTER_TO_INT(cmd) == -1) {
> + g_main_loop_quit(loop);
> + return FALSE;
> + }
> +
> + *ext = *cmd;
> +
> + return TRUE;
> +}
> +
> +static int req_cmd_notification(QXLInstance *qin)
> +{
> + if (!started)
> + return TRUE;
> +
> + spice_printerr("id: %d, queue length: %d",
> + fill_source_id, g_async_queue_length(aqueue));
I'd personally prefer to use glib functions in new code: ->
g_printerr()?
> +
> + return TRUE;
> +}
> +
> +static void end_replay(void)
> +{
> + int child_status;
> +
> + /* FIXME: wait threads and end cleanly */
> + spice_replay_free(replay);
> +
> + if (client_pid) {
> + g_debug("kill %d", client_pid);
> + kill(client_pid, SIGINT);
> + waitpid(client_pid, &child_status, 0);
> + }
> + exit(0);
> +}
> +
> +static void release_resource(QXLInstance *qin, struct QXLReleaseInfoExt release_info)
> +{
> + spice_replay_free_cmd(replay, (QXLCommandExt *)release_info.info->id);
> +}
> +
> +static int get_cursor_command(QXLInstance *qin, struct QXLCommandExt *ext)
> +{
> + return FALSE;
> +}
> +
> +static int req_cursor_notification(QXLInstance *qin)
> +{
> + return TRUE;
> +}
> +
> +static void notify_update(QXLInstance *qin, uint32_t update_id)
> +{
> +}
> +
> +static int flush_resources(QXLInstance *qin)
> +{
> + return TRUE;
> +}
> +
> +static QXLInterface display_sif = {
> + .base = {
> + .type = SPICE_INTERFACE_QXL,
> + .description = "replay",
> + .major_version = SPICE_INTERFACE_QXL_MAJOR,
> + .minor_version = SPICE_INTERFACE_QXL_MINOR
> + },
> + .attache_worker = attach_worker,
> + .set_compression_level = set_compression_level,
> + .set_mm_time = set_mm_time,
> + .get_init_info = get_init_info,
> + .get_command = get_command,
> + .req_cmd_notification = req_cmd_notification,
> + .release_resource = release_resource,
> + .get_cursor_command = get_cursor_command,
> + .req_cursor_notification = req_cursor_notification,
> + .notify_update = notify_update,
> + .flush_resources = flush_resources,
> +};
> +
> +static void replay_channel_event(int event, SpiceChannelEventInfo *info)
> +{
> + spice_printerr("");
g_printerr()
> +
> + if (info->type == SPICE_CHANNEL_DISPLAY &&
> + event == SPICE_CHANNEL_EVENT_INITIALIZED) {
> + started = TRUE;
> + }
> +}
> +
> +static gboolean start_client(gchar *cmd, GError **error)
> +{
> + gboolean retval;
> + gint argc;
> + gchar **argv = NULL;
> +
> +
> + if (!g_shell_parse_argv(cmd, &argc, &argv, error))
> + return FALSE;
> +
> + retval = g_spawn_async(NULL, argv, NULL, G_SPAWN_SEARCH_PATH,
> + NULL, NULL, &client_pid, error);
> + g_strfreev(argv);
> +
> + return retval;
> +}
> +
> +static gboolean progress_timer(gpointer user_data)
> +{
> + FILE *fd = user_data;
> + /* it seems somehow thread safe, move to worker thread? */
> + double pos = (double)ftell(fd);
> +
> + g_debug("%.2f%%", pos/total_size * 100);
> + return TRUE;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + GError *error = NULL;
> + GOptionContext *context = NULL;
> + gchar *client = NULL, **file = NULL;
> + gint port = 5000, compression = SPICE_IMAGE_COMPRESSION_AUTO_GLZ;
> + gboolean wait = FALSE;
> + FILE *fd;
> +
> + GOptionEntry entries[] = {
> + { "client", 'c', 0, G_OPTION_ARG_STRING, &client, "Client", "CMD" },
> + { "compression", 'C', 0, G_OPTION_ARG_INT, &compression, "Compression (default 2)", "INT" },
> + { "port", 'p', 0, G_OPTION_ARG_INT, &port, "Server port (default 5000)", "PORT" },
> + { "wait", 'w', 0, G_OPTION_ARG_NONE, &wait, "Wait for client", NULL },
> + { "slow", 's', 0, G_OPTION_ARG_INT, &slow, "Slow down replay", NULL },
> + { G_OPTION_REMAINING, 0, 0, G_OPTION_ARG_FILENAME_ARRAY, &file, "replay file", "FILE" },
This last option specifies an array of files, but the code below only
opens the first one.
> + { NULL }
> + };
> +
> + context = g_option_context_new("- replay spice server recording");
This is a common use of this "parameter_string" argument (and it's even
mentioned as an alternative in the API documentation), but I really
dislike it and find it confusing. I would prefer to use
g_option_context_set_summary() for the description.
> + g_option_context_add_main_entries(context, entries, NULL);
> + if (!g_option_context_parse(context, &argc, &argv, &error)) {
> + g_printerr("Option parsing failed: %s\n", error->message);
> + exit(1);
> + }
> + if (!file) {
> + g_printerr(g_option_context_get_help(context, TRUE, NULL));
> + exit(1);
> + }
> +
> + if (strncmp(file[0], "-", 1) == 0) {
> + fd = stdin;
> + } else {
> + fd = fopen(file[0], "r");
> + }
> + if (fd == NULL) {
> + g_printerr("error opening %s\n", argv[1]);
> + return 1;
> + }
> + if (fcntl(fileno(fd), FD_CLOEXEC) < 0) {
> + perror("fcntl failed");
> + exit(1);
> + }
> + fseek(fd, 0L, SEEK_END);
> + total_size = ftell(fd);
> + fseek(fd, 0L, SEEK_SET);
> + if (total_size > 0)
> + g_timeout_add_seconds(1, progress_timer, fd);
> + replay = spice_replay_new(fd, MAX_SURFACE_NUM);
> +
> + aqueue = g_async_queue_new();
> + core = basic_event_loop_init();
> + core->channel_event = replay_channel_event;
> +
> + SpiceServer* server = spice_server_new();
> + spice_server_set_image_compression(server, compression);
> + spice_server_set_port(server, port);
> + spice_server_set_noauth(server);
> +
> + g_print("listening on port %d (unsecure)\n", port);
"unsecure" should be "insecure"
> + spice_server_init(server, core);
> +
> + display_sin.base.sif = &display_sif.base;
> + spice_server_add_interface(server, &display_sin.base);
> +
> + if (client) {
> + start_client(client, &error);
> + wait = TRUE;
> + }
> +
> + if (!wait) {
> + started = TRUE;
> + fill_queue();
> + }
> +
> + loop = g_main_loop_new(NULL, FALSE);
> + g_main_loop_run(loop);
> +
> + end_replay();
> + g_async_queue_unref(aqueue);
> +
> + /* FIXME: there should be a way to join server threads before:
> + * g_main_loop_unref(loop);
> + */
> +
> + return 0;
> +}
More information about the Spice-devel
mailing list