qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH v1 08/22] migration: convert post-copy to use QI


From: Dr. David Alan Gilbert
Subject: Re: [Qemu-devel] [PATCH v1 08/22] migration: convert post-copy to use QIOChannelBuffer
Date: Tue, 26 Jan 2016 18:59:29 +0000
User-agent: Mutt/1.5.24 (2015-08-30)

* Daniel P. Berrange (address@hidden) wrote:
> On Mon, Jan 25, 2016 at 07:38:59PM +0000, Dr. David Alan Gilbert wrote:
> > * Daniel P. Berrange (address@hidden) wrote:
> > > The post-copy code does some I/O to/from an intermediate
> > > in-memory buffer rather than direct to the underlying
> > > I/O channel. Switch this code to use QIOChannelBuffer
> > > instead of QEMUSizedBuffer.
> > > 
> > > Signed-off-by: Daniel P. Berrange <address@hidden>
> > > ---
> > >  docs/migration.txt      |  4 ++--
> > >  include/sysemu/sysemu.h |  2 +-
> > >  migration/migration.c   | 15 +++++++--------
> > >  migration/savevm.c      | 41 +++++++++++++----------------------------
> > >  4 files changed, 23 insertions(+), 39 deletions(-)
> > > 
> > > diff --git a/docs/migration.txt b/docs/migration.txt
> > > index fda8d61..11703de 100644
> > > --- a/docs/migration.txt
> > > +++ b/docs/migration.txt
> > > @@ -403,8 +403,8 @@ listen thread:                     --- page -- page 
> > > -- page -- page -- page --
> > >  
> > >  On receipt of CMD_PACKAGED (1)
> > >     All the data associated with the package - the ( ... ) section in the
> > > -diagram - is read into memory (into a QEMUSizedBuffer), and the main 
> > > thread
> > > -recurses into qemu_loadvm_state_main to process the contents of the 
> > > package (2)
> > > +diagram - is read into memory, and the main thread recurses into
> > > +qemu_loadvm_state_main to process the contents of the package (2)
> > >  which contains commands (3,6) and devices (4...)
> > >  
> > >  On receipt of 'postcopy listen' - 3 -(i.e. the 1st command in the 
> > > package)
> > > diff --git a/include/sysemu/sysemu.h b/include/sysemu/sysemu.h
> > > index 3bb8897..0f59912 100644
> > > --- a/include/sysemu/sysemu.h
> > > +++ b/include/sysemu/sysemu.h
> > > @@ -120,7 +120,7 @@ void qemu_savevm_command_send(QEMUFile *f, enum 
> > > qemu_vm_cmd command,
> > >                                uint16_t len, uint8_t *data);
> > >  void qemu_savevm_send_ping(QEMUFile *f, uint32_t value);
> > >  void qemu_savevm_send_open_return_path(QEMUFile *f);
> > > -int qemu_savevm_send_packaged(QEMUFile *f, const QEMUSizedBuffer *qsb);
> > > +int qemu_savevm_send_packaged(QEMUFile *f, const char *buf, size_t len);
> > >  void qemu_savevm_send_postcopy_advise(QEMUFile *f);
> > >  void qemu_savevm_send_postcopy_listen(QEMUFile *f);
> > >  void qemu_savevm_send_postcopy_run(QEMUFile *f);
> > > diff --git a/migration/migration.c b/migration/migration.c
> > > index 715f069..e921b20 100644
> > > --- a/migration/migration.c
> > > +++ b/migration/migration.c
> > > @@ -33,6 +33,7 @@
> > >  #include "qom/cpu.h"
> > >  #include "exec/memory.h"
> > >  #include "exec/address-spaces.h"
> > > +#include "io/channel-buffer.h"
> > >  
> > >  #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed 
> > > throttling */
> > >  
> > > @@ -1401,7 +1402,8 @@ static int 
> > > await_return_path_close_on_source(MigrationState *ms)
> > >  static int postcopy_start(MigrationState *ms, bool *old_vm_running)
> > >  {
> > >      int ret;
> > > -    const QEMUSizedBuffer *qsb;
> > > +    QIOChannelBuffer *bioc;
> > > +    QEMUFile *fb;
> > >      int64_t time_at_stop = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
> > >      migrate_set_state(ms, MIGRATION_STATUS_ACTIVE,
> > >                        MIGRATION_STATUS_POSTCOPY_ACTIVE);
> > > @@ -1456,11 +1458,9 @@ static int postcopy_start(MigrationState *ms, bool 
> > > *old_vm_running)
> > >       * So we wrap the device state up in a package with a length at the 
> > > start;
> > >       * to do this we use a qemu_buf to hold the whole of the device 
> > > state.
> > >       */
> > > -    QEMUFile *fb = qemu_bufopen("w", NULL);
> > > -    if (!fb) {
> > > -        error_report("Failed to create buffered file");
> > > -        goto fail;
> > > -    }
> > > +    bioc = qio_channel_buffer_new(4096);
> > > +    fb = qemu_fopen_channel_input(QIO_CHANNEL(bioc));
> > 
> > Why is that _input ?
> 
> A mistake - should be output given that the old code used "w".

OK, that makes sense then.

> > Also, what's your naming converntion; my 'qsb' was just an
> > abbreviation of (Q)EMU(S)ized(B)uffer;  what's a bioc?
> 
> In the vast majority of cases where you just have a pointer
> to the generic base class QIOChannel, then I'll use 'ioc' as
> the standard variable name - ie 'io' and 'c' for channel.
> 
> If I have a specialized subtype, I prefix a single letter
> reflecting the name of the subtype. eg 's' for QIOChannelSocket
> giving "sioc" and 'b' for QIOChannelBuffer giving "bioc" etc.

OK, my preference there would have been iocb just because it's the
same order as the thing it's abbreviating; but that's just a personal
preference.

> > > +    object_unref(OBJECT(bioc));
> > >  
> > >      /*
> > >       * Make sure the receiver can get incoming pages before we send the 
> > > rest
> > > @@ -1474,10 +1474,9 @@ static int postcopy_start(MigrationState *ms, bool 
> > > *old_vm_running)
> > >      qemu_savevm_send_postcopy_run(fb);
> > >  
> > >      /* <><> end of stuff going into the package */
> > > -    qsb = qemu_buf_get(fb);
> > >  
> > >      /* Now send that blob */
> > > -    if (qemu_savevm_send_packaged(ms->file, qsb)) {
> > > +    if (qemu_savevm_send_packaged(ms->file, bioc->data, bioc->usage)) {
> > >          goto fail_closefb;
> > >      }
> > >      qemu_fclose(fb);
> > > diff --git a/migration/savevm.c b/migration/savevm.c
> > > index 0ad1b93..f2e1880 100644
> > > --- a/migration/savevm.c
> > > +++ b/migration/savevm.c
> > > @@ -50,6 +50,7 @@
> > >  #include "qemu/iov.h"
> > >  #include "block/snapshot.h"
> > >  #include "block/qapi.h"
> > > +#include "io/channel-buffer.h"
> > >  
> > >  
> > >  #ifndef ETH_P_RARP
> > > @@ -760,10 +761,8 @@ void qemu_savevm_send_open_return_path(QEMUFile *f)
> > >   *    0 on success
> > >   *    -ve on error
> > >   */
> > > -int qemu_savevm_send_packaged(QEMUFile *f, const QEMUSizedBuffer *qsb)
> > > +int qemu_savevm_send_packaged(QEMUFile *f, const char *buf, size_t len)
> > >  {
> > > -    size_t cur_iov;
> > > -    size_t len = qsb_get_length(qsb);
> > >      uint32_t tmp;
> > >  
> > >      if (len > MAX_VM_CMD_PACKAGED_SIZE) {
> > > @@ -777,18 +776,7 @@ int qemu_savevm_send_packaged(QEMUFile *f, const 
> > > QEMUSizedBuffer *qsb)
> > >      trace_qemu_savevm_send_packaged();
> > >      qemu_savevm_command_send(f, MIG_CMD_PACKAGED, 4, (uint8_t *)&tmp);
> > >  
> > > -    /* all the data follows (concatinating the iov's) */
> > > -    for (cur_iov = 0; cur_iov < qsb->n_iov; cur_iov++) {
> > > -        /* The iov entries are partially filled */
> > > -        size_t towrite = MIN(qsb->iov[cur_iov].iov_len, len);
> > > -        len -= towrite;
> > > -
> > > -        if (!towrite) {
> > > -            break;
> > > -        }
> > > -
> > > -        qemu_put_buffer(f, qsb->iov[cur_iov].iov_base, towrite);
> > > -    }
> > > +    qemu_put_buffer(f, (const uint8_t *)buf, len);
> > >  
> > >      return 0;
> > >  }
> > > @@ -1548,9 +1536,8 @@ static int 
> > > loadvm_postcopy_handle_run(MigrationIncomingState *mis)
> > >  static int loadvm_handle_cmd_packaged(MigrationIncomingState *mis)
> > >  {
> > >      int ret;
> > > -    uint8_t *buffer;
> > >      uint32_t length;
> > > -    QEMUSizedBuffer *qsb;
> > > +    QIOChannelBuffer *bioc;
> > >  
> > >      length = qemu_get_be32(mis->from_src_file);
> > >      trace_loadvm_handle_cmd_packaged(length);
> > > @@ -1559,28 +1546,26 @@ static int 
> > > loadvm_handle_cmd_packaged(MigrationIncomingState *mis)
> > >          error_report("Unreasonably large packaged state: %u", length);
> > >          return -1;
> > >      }
> > > -    buffer = g_malloc0(length);
> > > -    ret = qemu_get_buffer(mis->from_src_file, buffer, (int)length);
> > > +
> > > +    bioc = qio_channel_buffer_new(length);
> > > +    ret = qemu_get_buffer(mis->from_src_file,
> > > +                          (uint8_t *)bioc->data,
> > > +                          (int)length);
> > 
> > Actually, qemu_get_buffer takes a size_t now; so you can fix that at the 
> > same time.
> 
> Great.
> 
> > 
> > >      if (ret != length) {
> > > -        g_free(buffer);
> > > +        object_unref(OBJECT(bioc));
> > >          error_report("CMD_PACKAGED: Buffer receive fail ret=%d 
> > > length=%d\n",
> > >                  ret, length);
> > >          return (ret < 0) ? ret : -EAGAIN;
> > >      }
> > > +    bioc->usage += length;
> > 
> > I'm a bit surprised you've not wrapped feeding the data
> > into the qio_channel_buffer up a bit rather than having the callers
> > prod at it's buffer/usage elements directly.
> 
> I figured it was better to avoid extra memcpy's in the migration code,
> so decided to directly pass the internal buffer to get_qemu_buffer().

Yes, avoiding copies is good; I was just thinking more of a function that
incremented the usage or something like that.

Dave

> 
> 
> Regards,
> Daniel
> -- 
> |: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
> |: http://libvirt.org              -o-             http://virt-manager.org :|
> |: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
> |: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK



reply via email to

[Prev in Thread] Current Thread [Next in Thread]