[Spice-devel] [PATCH v7 spice-gtk] Usbredir: enable lz4 compression

Snir Sheriber ssheribe at redhat.com
Mon Jun 27 14:54:37 UTC 2016


Hi,

On 06/27/2016 04:44 PM, Victor Toso wrote:
> Hi,
>
> On Sun, Jun 26, 2016 at 04:44:42PM +0300, Snir Sheriber wrote:
>> Hi,
>>
>> On 06/24/2016 06:57 PM, Victor Toso wrote:
>>> Hi,
>>>
>>> On Mon, Jun 13, 2016 at 07:54:34PM +0300, Snir Sheriber wrote:
>>>> Compressed message type is CompressedData which contains compression
>>>> type (1 byte) followed by the uncompressed data size (4 bytes-exists
>>>> only if data was compressed) followed by the compressed data
>>>>
>>>> If SPICE_SPICEVMC_CAP_DATA_COMPRESS_LZ4 capability is available &&
>>>> data_size > COMPRESS_THRESHOLD && !AF_LOCAL data will be sent
>>>> compressed otherwise data will be sent uncompressed (as well if
>>>> compression has failed)
>>>>
>>>> Update required spice-protocol to 0.12.12
>>> Now that we did the spice-gtk release, updating the spice-common and
>>> bumping the spice-protocol requirement should be okay.
>>>
>>> I tested this before but I plan to double check it today, if no
>>> problems and no one complains, I plan to push this next by Monday.
>>>
>>> Minor comments below:
>>>
>>>> Signed-off-by: Snir Sheriber <ssheribe at redhat.com>
>>>> ---
>>>>    configure.ac           |   2 +-
>>>>    src/channel-usbredir.c | 129 ++++++++++++++++++++++++++++++++++++++++++++++---
>>>>    2 files changed, 124 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/configure.ac b/configure.ac
>>>> index 3fe8055..f2acb23 100644
>>>> --- a/configure.ac
>>>> +++ b/configure.ac
>>>> @@ -69,7 +69,7 @@ AC_CHECK_LIBM
>>>>    AC_SUBST(LIBM)
>>>>    AC_CONFIG_SUBDIRS([spice-common])
>>>> -PKG_CHECK_MODULES([SPICE_PROTOCOL], [spice-protocol >= 0.12.11])
>>>> +PKG_CHECK_MODULES([SPICE_PROTOCOL], [spice-protocol >= 0.12.12])
>>>>    COMMON_CFLAGS='-I ${top_srcdir}/spice-common/ ${SPICE_PROTOCOL_CFLAGS}'
>>>>    AC_SUBST(COMMON_CFLAGS)
>>>> diff --git a/src/channel-usbredir.c b/src/channel-usbredir.c
>>>> index 2c5feae..bb69e26 100644
>>>> --- a/src/channel-usbredir.c
>>>> +++ b/src/channel-usbredir.c
>>>> @@ -24,6 +24,9 @@
>>>>    #ifdef USE_USBREDIR
>>>>    #include <glib/gi18n.h>
>>>>    #include <usbredirhost.h>
>>>> +#ifdef USE_LZ4
>>>> +#include <lz4.h>
>>>> +#endif
>>>>    #ifdef USE_POLKIT
>>>>    #include "usb-acl-helper.h"
>>>>    #endif
>>>> @@ -32,6 +35,7 @@
>>>>    #include "usbutil.h"
>>>>    #endif
>>>> +#include "common/log.h"
>>>>    #include "spice-client.h"
>>>>    #include "spice-common.h"
>>>> @@ -51,6 +55,7 @@
>>>>    #ifdef USE_USBREDIR
>>>> +#define COMPRESS_THRESHOLD 1000
>>>>    #define SPICE_USBREDIR_CHANNEL_GET_PRIVATE(obj)                                  \
>>>>        (G_TYPE_INSTANCE_GET_PRIVATE((obj), SPICE_TYPE_USBREDIR_CHANNEL, SpiceUsbredirChannelPrivate))
>>>> @@ -233,6 +238,7 @@ static void channel_set_handlers(SpiceChannelClass *klass)
>>>>    {
>>>>        static const spice_msg_handler handlers[] = {
>>>>            [ SPICE_MSG_SPICEVMC_DATA ] = usbredir_handle_msg,
>>>> +        [ SPICE_MSG_SPICEVMC_COMPRESSED_DATA ] = usbredir_handle_msg,
>>>>        };
>>>>        spice_channel_set_handlers(klass, handlers, G_N_ELEMENTS(handlers));
>>>> @@ -269,6 +275,9 @@ void spice_usbredir_channel_set_context(SpiceUsbredirChannel *channel,
>>>>    #if USBREDIR_VERSION >= 0x000701
>>>>        usbredirhost_set_buffered_output_size_cb(priv->host, usbredir_buffered_output_size_callback);
>>>>    #endif
>>>> +#ifdef USE_LZ4
>>>> +    spice_channel_set_capability(channel, SPICE_SPICEVMC_CAP_DATA_COMPRESS_LZ4);
>>>> +#endif
>>>>    }
>>>>    static gboolean spice_usbredir_channel_open_device(
>>>> @@ -632,11 +641,71 @@ static void usbredir_free_write_cb_data(uint8_t *data, void *user_data)
>>>>        usbredirhost_free_write_buffer(priv->host, data);
>>>>    }
>>>> +#ifdef USE_LZ4
>>>> +static int try_write_compress_LZ4(SpiceUsbredirChannel *channel, uint8_t *data, int count) {
>>>> +    SpiceChannelPrivate *c;
>>>> +    SpiceMsgOut *msg_out_compressed;
>>>> +    int bound, compressed_data_count;
>>>> +    uint8_t *compressed_buf;
>>>> +    SpiceMsgCompressedData compressed_data_msg = {
>>>> +            .type = SPICE_DATA_COMPRESSION_TYPE_LZ4,
>>>> +            .uncompressed_size = count
>>>> +        };
>>>> +
>>>> +    c = SPICE_CHANNEL(channel)->priv;
>>>> +    if (g_socket_get_family(c->sock) == G_SOCKET_FAMILY_UNIX) {
>>>> +        /* AF_LOCAL socket - data will not be compressed */
>>>> +        return FALSE;
>>>> +    }
>>>> +    if (count <= COMPRESS_THRESHOLD) {
>>>> +        /* Not enough data to compress */
>>>> +        return FALSE;
>>>> +    }
>>>> +    if (!spice_channel_test_capability(SPICE_CHANNEL(channel),
>>>> +                                       SPICE_SPICEVMC_CAP_DATA_COMPRESS_LZ4)) {
>>>> +        /* No server compression capability - data will not be compressed */
>>>> +        return FALSE;
>>>> +    }
>>>> +    bound = LZ4_compressBound(count);
>>>> +    if (bound == 0) {
>>>> +        /* Invalid bound - data will not be compressed */
>>>> +        return FALSE;
>>>> +    }
>>> What about devices of isochronous type? Here, the packages for my webcam
>>> were < COMPRESS_THRESHOLD but I guess we could save some time/process in
>>> general by avoiding compressing in case the device is isoc.
>> yes, it should be better to send data of isochronous devices uncompressed
>> (although it is not guaranteed that isoc devices data is already
>> compressed).
>> as far as i looked with Uri there are two ways to do so:
>> 1. by looking for the type in each of the endpoints descriptors via
>> libusb
>> 2. to check if usbredir_buffered_output_size_callback is being called
>> and if it is, it means the device is isoc (kind of hackish)
>> what do you think should be used here?
> (1) should be better indeed. The problem with (2) is that the callback
> might not be called unless there are data to be dropped.
>
> For me, this could (should?) be an additional patch but let me know if
> you want me to have them upstream together otherwise I don't really mind
> to have this as it is (with fixes for build, but I can do that)
>
> Cheers,
>    toso
>
Thanks!
Sure, in additional patch it will be better.
(and i can also do this build-fix & rebase if needed)

Snir.
>>>> +
>>>> +    compressed_buf = (uint8_t*)spice_malloc(bound);
>>>> +    compressed_data_count = LZ4_compress_default((char*)data,
>>>> +                                                 (char*)compressed_buf,
>>>> +                                                 count,
>>>> +                                                 bound);
>>>> +    if (compressed_data_count > 0 && compressed_data_count < count) {
>>>> +        compressed_data_msg.compressed_data = compressed_buf;
>>>> +        msg_out_compressed = spice_msg_out_new(SPICE_CHANNEL(channel),
>>>> +                                               SPICE_MSGC_SPICEVMC_COMPRESSED_DATA);
>>>> +        msg_out_compressed->marshallers->msg_SpiceMsgCompressedData(msg_out_compressed->marshaller,
>>>> +                                                                    &compressed_data_msg);
>>>> +        spice_marshaller_add_ref_full(msg_out_compressed->marshaller,
>>>> +                                      compressed_data_msg.compressed_data,
>>>> +                                      compressed_data_count,
>>>> +                                      usbredir_free_write_cb_data,
>>>> +                                      channel);
>>>> +        spice_msg_out_send(msg_out_compressed);
>>>> +        return TRUE;
>>>> +    }
>>>> +    /* if not - free & fallback to sending the message uncompressed */
>>>> +    free(compressed_buf);
>>>> +    return FALSE;
>>>> +}
>>>> +#endif
>>>> +
>>>>    static int usbredir_write_callback(void *user_data, uint8_t *data, int count)
>>>>    {
>>>>        SpiceUsbredirChannel *channel = user_data;
>>>>        SpiceMsgOut *msg_out;
>>>> +#ifdef USE_LZ4
>>>> +    if (try_write_compress_LZ4(channel, data, count))
>>>> +        return count;
>>>> +#endif
>>>>        msg_out = spice_msg_out_new(SPICE_CHANNEL(channel),
>>>>                                    SPICE_MSGC_SPICEVMC_DATA);
>>>>        spice_marshaller_add_ref_full(msg_out->marshaller, data, count,
>>>> @@ -724,11 +793,49 @@ static void spice_usbredir_channel_up(SpiceChannel *c)
>>>>        usbredirhost_write_guest_data(priv->host);
>>>>    }
>>>> +static int try_handle_compressed_msg(SpiceMsgCompressedData *compressed_data_msg,
>>>> +                                     uint8_t **buf,
>>>> +                                     int *size) {
>>>> +    int decompressed_size = 0;
>>>> +    char *decompressed = NULL;
>>>> +
>>>> +    if (compressed_data_msg->uncompressed_size == 0) {
>>>> +        spice_warning("Invalid uncompressed_size");
>>>> +        return FALSE;
>>>> +    }
>>>> +
>>>> +    switch (compressed_data_msg->type) {
>>>> +#ifdef USE_LZ4
>>>> +    case SPICE_DATA_COMPRESSION_TYPE_LZ4:
>>>> +        decompressed = g_malloc(compressed_data_msg->uncompressed_size);
>>>> +        decompressed_size = LZ4_decompress_safe ((char*)compressed_data_msg->compressed_data,
>>>> +                                                 decompressed,
>>>> +                                                 compressed_data_msg->compressed_size,
>>>> +                                                 compressed_data_msg->uncompressed_size);
>>>> +        break;
>>>> +#endif
>>>> +    default:
>>>> +        spice_warning("Unknown Compression Type");
>>>> +        return FALSE;
>>>> +    }
>>>> +    if (decompressed_size != compressed_data_msg->uncompressed_size) {
>>>> +        spice_warning("Decompress Error decompressed_size=%d expected=%d",
>>>> +                      decompressed_size, compressed_data_msg->uncompressed_size);
>>> with -Werror-format, build fails here with:
>>>    error: format ‘%d’ expects argument of type ‘int’, but argument 7 has
>>>    type ‘uint32_t {aka unsigned int}’ expected=%d",
>>>
>>> Cheers,
>>>     toso
>> Thanks! shouldn't be missed :/
>>>> +        g_free(decompressed);
>>>> +        return FALSE;
>>>> +    }
>>>> +
>>>> +    *size = decompressed_size;
>>>> +    *buf = (uint8_t*)decompressed;
>>>> +    return TRUE;
>>>> +
>>>> +}
>>>> +
>>>>    static void usbredir_handle_msg(SpiceChannel *c, SpiceMsgIn *in)
>>>>    {
>>>>        SpiceUsbredirChannel *channel = SPICE_USBREDIR_CHANNEL(c);
>>>>        SpiceUsbredirChannelPrivate *priv = channel->priv;
>>>> -    int r, size;
>>>> +    int r = 0, size;
>>>>        uint8_t *buf;
>>>>        g_return_if_fail(priv->host != NULL);
>>>> @@ -736,13 +843,23 @@ static void usbredir_handle_msg(SpiceChannel *c, SpiceMsgIn *in)
>>>>        /* No recursion allowed! */
>>>>        g_return_if_fail(priv->read_buf == NULL);
>>>> -    buf = spice_msg_in_raw(in, &size);
>>>> -    priv->read_buf = buf;
>>>> -    priv->read_buf_size = size;
>>>> +    if (spice_msg_in_type(in) == SPICE_MSG_SPICEVMC_COMPRESSED_DATA) {
>>>> +        SpiceMsgCompressedData *compressed_data_msg = spice_msg_in_parsed(in);
>>>> +        if (try_handle_compressed_msg(compressed_data_msg, &buf, &size)) {
>>>> +            priv->read_buf_size = size;
>>>> +            priv->read_buf = buf;
>>>> +        } else {
>>>> +            r = usbredirhost_read_parse_error;
>>>> +        }
>>>> +    } else { /* Regular SPICE_MSG_SPICEVMC_DATA msg */
>>>> +        buf = spice_msg_in_raw(in, &size);
>>>> +        priv->read_buf_size = size;
>>>> +        priv->read_buf = buf;
>>>> +    }
>>>>        spice_usbredir_channel_lock(channel);
>>>> -
>>>> -    r = usbredirhost_read_guest_data(priv->host);
>>>> +    if (r == 0)
>>>> +        r = usbredirhost_read_guest_data(priv->host);
>>>>        if (r != 0) {
>>>>            SpiceUsbDevice *spice_device = priv->spice_device;
>>>>            device_error_data err_data;
>>>> -- 
>>>> 2.5.5
>>>>
>>>> _______________________________________________
>>>> Spice-devel mailing list
>>>> Spice-devel at lists.freedesktop.org
>>>> https://lists.freedesktop.org/mailman/listinfo/spice-devel



More information about the Spice-devel mailing list