qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Avihai Horon <avihaih@nvidia.com>
To: Peter Xu <peterx@redhat.com>
Cc: qemu-devel@nongnu.org,
	"Alex Williamson" <alex.williamson@redhat.com>,
	"Cédric Le Goater" <clg@redhat.com>,
	"Juan Quintela" <quintela@redhat.com>,
	"Leonardo Bras" <leobras@redhat.com>,
	"Eric Blake" <eblake@redhat.com>,
	"Markus Armbruster" <armbru@redhat.com>,
	"Thomas Huth" <thuth@redhat.com>,
	"Laurent Vivier" <lvivier@redhat.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Yishai Hadas" <yishaih@nvidia.com>,
	"Jason Gunthorpe" <jgg@nvidia.com>,
	"Maor Gottlieb" <maorg@nvidia.com>,
	"Kirti Wankhede" <kwankhede@nvidia.com>,
	"Tarun Gupta" <targupta@nvidia.com>,
	"Joao Martins" <joao.m.martins@oracle.com>
Subject: Re: [PATCH 2/8] migration: Add precopy initial data handshake
Date: Wed, 3 May 2023 18:31:40 +0300	[thread overview]
Message-ID: <dc5af4cd-1844-5b67-2eb1-56dbee5d9dec@nvidia.com> (raw)
In-Reply-To: <ZFGUriGbwleLEIcm@x1n>


On 03/05/2023 1:54, Peter Xu wrote:
> External email: Use caution opening links or attachments
>
>
> (I've left high level comment in cover letter, but still some quick
> comments I noticed when reading)
>
> On Mon, May 01, 2023 at 05:01:35PM +0300, Avihai Horon wrote:
>> Add precopy initial data handshake between source and destination upon
>> migration setup. The purpose of the handshake is to notify the
>> destination that precopy initial data is used and which migration users
>> (i.e., SaveStateEntry) are going to use it.
>>
>> The handshake is done in two levels. First, a general enable command is
>> sent to notify the destination migration code that precopy initial data
>> is used.
>> Then, for each migration user in the source that supports precopy
>> initial data, an enable command is sent to its counterpart in the
>> destination:
>> If both support it, precopy initial data will be used for them.
>> If source doesn't support it, precopy initial data will not be used for
>> them.
>> If source supports it and destination doesn't, migration will be failed.
>>
>> To implement it, a new migration command MIG_CMD_INITIAL_DATA_ENABLE is
>> added, as well as a new SaveVMHandlers handler initial_data_advise.
>> Calling the handler advises the migration user that precopy initial data
>> is used and its return value indicates whether precopy initial data is
>> supported by it.
>>
>> Signed-off-by: Avihai Horon <avihaih@nvidia.com>
>> ---
>>   include/migration/register.h |   6 +++
>>   migration/migration.h        |   3 ++
>>   migration/savevm.h           |   1 +
>>   migration/migration.c        |   4 ++
>>   migration/savevm.c           | 102 +++++++++++++++++++++++++++++++++++
>>   migration/trace-events       |   2 +
>>   6 files changed, 118 insertions(+)
>>
>> diff --git a/include/migration/register.h b/include/migration/register.h
>> index a8dfd8fefd..0a73f3883e 100644
>> --- a/include/migration/register.h
>> +++ b/include/migration/register.h
>> @@ -71,6 +71,12 @@ typedef struct SaveVMHandlers {
>>       int (*load_cleanup)(void *opaque);
>>       /* Called when postcopy migration wants to resume from failure */
>>       int (*resume_prepare)(MigrationState *s, void *opaque);
>> +
>> +    /*
>> +     * Advises that precopy initial data was requested to be enabled. Returns
>> +     * true if it's supported or false otherwise. Called both in src and dest.
>> +     */
>> +    bool (*initial_data_advise)(void *opaque);
>>   } SaveVMHandlers;
>>
>>   int register_savevm_live(const char *idstr,
>> diff --git a/migration/migration.h b/migration/migration.h
>> index 3a918514e7..4f615e9dbc 100644
>> --- a/migration/migration.h
>> +++ b/migration/migration.h
>> @@ -204,6 +204,9 @@ struct MigrationIncomingState {
>>        * contains valid information.
>>        */
>>       QemuMutex page_request_mutex;
>> +
>> +    /* Indicates whether precopy initial data was enabled and should be used */
>> +    bool initial_data_enabled;
>>   };
>>
>>   MigrationIncomingState *migration_incoming_get_current(void);
>> diff --git a/migration/savevm.h b/migration/savevm.h
>> index fb636735f0..d47ab4ad18 100644
>> --- a/migration/savevm.h
>> +++ b/migration/savevm.h
>> @@ -58,6 +58,7 @@ void qemu_savevm_send_postcopy_ram_discard(QEMUFile *f, const char *name,
>>                                              uint64_t *start_list,
>>                                              uint64_t *length_list);
>>   void qemu_savevm_send_colo_enable(QEMUFile *f);
>> +void qemu_savevm_send_initial_data_enable(MigrationState *ms, QEMUFile *f);
>>   void qemu_savevm_live_state(QEMUFile *f);
>>   int qemu_save_device_state(QEMUFile *f);
>>
>> diff --git a/migration/migration.c b/migration/migration.c
>> index abcadbb619..68cdf5b184 100644
>> --- a/migration/migration.c
>> +++ b/migration/migration.c
>> @@ -2964,6 +2964,10 @@ static void *migration_thread(void *opaque)
>>           qemu_savevm_send_colo_enable(s->to_dst_file);
>>       }
>>
>> +    if (migrate_precopy_initial_data()) {
>> +        qemu_savevm_send_initial_data_enable(s, s->to_dst_file);
>> +    }
>> +
>>       qemu_savevm_state_setup(s->to_dst_file);
>>
>>       qemu_savevm_wait_unplug(s, MIGRATION_STATUS_SETUP,
>> diff --git a/migration/savevm.c b/migration/savevm.c
>> index a9181b444b..2740defdf0 100644
>> --- a/migration/savevm.c
>> +++ b/migration/savevm.c
>> @@ -71,6 +71,13 @@
>>
>>   const unsigned int postcopy_ram_discard_version;
>>
>> +typedef struct {
>> +    uint8_t general_enable;
>> +    uint8_t reserved[7];
>> +    uint8_t idstr[256];
>> +    uint32_t instance_id;
>> +} InitialDataInfo;
>> +
>>   /* Subcommands for QEMU_VM_COMMAND */
>>   enum qemu_vm_cmd {
>>       MIG_CMD_INVALID = 0,   /* Must be 0 */
>> @@ -90,6 +97,8 @@ enum qemu_vm_cmd {
>>       MIG_CMD_ENABLE_COLO,       /* Enable COLO */
>>       MIG_CMD_POSTCOPY_RESUME,   /* resume postcopy on dest */
>>       MIG_CMD_RECV_BITMAP,       /* Request for recved bitmap on dst */
>> +
>> +    MIG_CMD_INITIAL_DATA_ENABLE, /* Enable precopy initial data in dest */
>>       MIG_CMD_MAX
>>   };
>>
>> @@ -109,6 +118,8 @@ static struct mig_cmd_args {
>>       [MIG_CMD_POSTCOPY_RESUME]  = { .len =  0, .name = "POSTCOPY_RESUME" },
>>       [MIG_CMD_PACKAGED]         = { .len =  4, .name = "PACKAGED" },
>>       [MIG_CMD_RECV_BITMAP]      = { .len = -1, .name = "RECV_BITMAP" },
>> +    [MIG_CMD_INITIAL_DATA_ENABLE] = { .len = sizeof(InitialDataInfo),
>> +                                      .name = "INITIAL_DATA_ENABLE" },
>>       [MIG_CMD_MAX]              = { .len = -1, .name = "MAX" },
>>   };
>>
>> @@ -1036,6 +1047,40 @@ static void qemu_savevm_command_send(QEMUFile *f,
>>       qemu_fflush(f);
>>   }
>>
>> +void qemu_savevm_send_initial_data_enable(MigrationState *ms, QEMUFile *f)
>> +{
>> +    SaveStateEntry *se;
>> +    InitialDataInfo buf;
>> +
>> +    /* Enable precopy initial data generally in the migration */
>> +    memset(&buf, 0, sizeof(buf));
>> +    buf.general_enable = 1;
>> +    qemu_savevm_command_send(f, MIG_CMD_INITIAL_DATA_ENABLE, sizeof(buf),
>> +                             (uint8_t *)&buf);
>> +    trace_savevm_send_initial_data_enable(buf.general_enable, (char *)buf.idstr,
>> +                                          buf.instance_id);
> Maybe a generalized helper would be nice here to send one
> MIG_CMD_INITIAL_DATA_ENABLE packet, then it can be used below too.

Sure, I will do that.

>
> Here instance_id is multi-bytes, we may want to consider endianess.  it
> seems the common way is use big endian over the wire for qemu migration
> msgs.

Oh, right. I will fix it.

Thanks.

>> +
>> +    /* Enable precopy initial data for each migration user that supports it */
>> +    QTAILQ_FOREACH(se, &savevm_state.handlers, entry) {
>> +        if (!se->ops || !se->ops->initial_data_advise) {
>> +            continue;
>> +        }
>> +
>> +        if (!se->ops->initial_data_advise(se->opaque)) {
>> +            continue;
>> +        }
>> +
>> +        memset(&buf, 0, sizeof(buf));
>> +        memcpy(buf.idstr, se->idstr, sizeof(buf.idstr));
>> +        buf.instance_id = se->instance_id;
>> +
>> +        qemu_savevm_command_send(f, MIG_CMD_INITIAL_DATA_ENABLE, sizeof(buf),
>> +                                 (uint8_t *)&buf);
>> +        trace_savevm_send_initial_data_enable(
>> +            buf.general_enable, (char *)buf.idstr, buf.instance_id);
>> +    }
>> +}
>> +
>>   void qemu_savevm_send_colo_enable(QEMUFile *f)
>>   {
>>       trace_savevm_send_colo_enable();
>> @@ -2314,6 +2359,60 @@ static int loadvm_process_enable_colo(MigrationIncomingState *mis)
>>       return ret;
>>   }
>>
>> +static int loadvm_handle_initial_data_enable(MigrationIncomingState *mis)
>> +{
>> +    InitialDataInfo buf;
>> +    SaveStateEntry *se;
>> +    ssize_t read_size;
>> +
>> +    read_size = qemu_get_buffer(mis->from_src_file, (void *)&buf, sizeof(buf));
>> +    if (read_size != sizeof(buf)) {
>> +        error_report("%s: Could not get data buffer, read_size %ld, len %lu",
>> +                     __func__, read_size, sizeof(buf));
>> +
>> +        return -EIO;
>> +    }
>> +
>> +    /* Enable precopy initial data generally in the migration */
>> +    if (buf.general_enable) {
>> +        mis->initial_data_enabled = true;
>> +        trace_loadvm_handle_initial_data_enable(
>> +            buf.general_enable, (char *)buf.idstr, buf.instance_id);
>> +
>> +        return 0;
>> +    }
>> +
>> +    /* Enable precopy initial data for a specific migration user */
>> +    se = find_se((char *)buf.idstr, buf.instance_id);
>> +    if (!se) {
>> +        error_report("%s: Could not find SaveStateEntry, idstr '%s', "
>> +                     "instance_id %" PRIu32,
>> +                     __func__, buf.idstr, buf.instance_id);
>> +
>> +        return -ENOENT;
>> +    }
>> +
>> +    if (!se->ops || !se->ops->initial_data_advise) {
>> +        error_report("%s: '%s' doesn't have required "
>> +                     "initial_data_advise op",
>> +                     __func__, buf.idstr);
>> +
>> +        return -EOPNOTSUPP;
>> +    }
>> +
>> +    if (!se->ops->initial_data_advise(se->opaque)) {
>> +        error_report("%s: '%s' doesn't support precopy initial data", __func__,
>> +                     buf.idstr);
>> +
>> +        return -EOPNOTSUPP;
>> +    }
>> +
>> +    trace_loadvm_handle_initial_data_enable(buf.general_enable,
>> +                                            (char *)buf.idstr, buf.instance_id);
>> +
>> +    return 0;
>> +}
>> +
>>   /*
>>    * Process an incoming 'QEMU_VM_COMMAND'
>>    * 0           just a normal return
>> @@ -2397,6 +2496,9 @@ static int loadvm_process_command(QEMUFile *f)
>>
>>       case MIG_CMD_ENABLE_COLO:
>>           return loadvm_process_enable_colo(mis);
>> +
>> +    case MIG_CMD_INITIAL_DATA_ENABLE:
>> +        return loadvm_handle_initial_data_enable(mis);
>>       }
>>
>>       return 0;
>> diff --git a/migration/trace-events b/migration/trace-events
>> index 92161eeac5..21ae471126 100644
>> --- a/migration/trace-events
>> +++ b/migration/trace-events
>> @@ -23,6 +23,7 @@ loadvm_postcopy_ram_handle_discard_end(void) ""
>>   loadvm_postcopy_ram_handle_discard_header(const char *ramid, uint16_t len) "%s: %ud"
>>   loadvm_process_command(const char *s, uint16_t len) "com=%s len=%d"
>>   loadvm_process_command_ping(uint32_t val) "0x%x"
>> +loadvm_handle_initial_data_enable(uint8_t general_enable, const char *idstr, int instance_id) "general_enable=%u, idstr=%s, instance_id=%u"
>>   postcopy_ram_listen_thread_exit(void) ""
>>   postcopy_ram_listen_thread_start(void) ""
>>   qemu_savevm_send_postcopy_advise(void) ""
>> @@ -38,6 +39,7 @@ savevm_send_postcopy_run(void) ""
>>   savevm_send_postcopy_resume(void) ""
>>   savevm_send_colo_enable(void) ""
>>   savevm_send_recv_bitmap(char *name) "%s"
>> +savevm_send_initial_data_enable(uint8_t general_enable, const char *idstr, int instance_id) "general_enable=%u, idstr=%s, instance_id=%u"
>>   savevm_state_setup(void) ""
>>   savevm_state_resume_prepare(void) ""
>>   savevm_state_header(void) ""
>> --
>> 2.26.3
>>
> --
> Peter Xu
>


  reply	other threads:[~2023-05-03 15:32 UTC|newest]

Thread overview: 42+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-01 14:01 [PATCH 0/8] migration: Add precopy initial data capability and VFIO precopy support Avihai Horon
2023-05-01 14:01 ` [PATCH 1/8] migration: Add precopy initial data capability Avihai Horon
2023-05-10  8:24   ` Juan Quintela
2023-05-17  9:17   ` Markus Armbruster
2023-05-17 10:16     ` Avihai Horon
2023-05-17 12:21       ` Markus Armbruster
2023-05-17 13:23         ` Avihai Horon
2023-05-01 14:01 ` [PATCH 2/8] migration: Add precopy initial data handshake Avihai Horon
2023-05-02 22:54   ` Peter Xu
2023-05-03 15:31     ` Avihai Horon [this message]
2023-05-10  8:40   ` Juan Quintela
2023-05-10 15:32     ` Avihai Horon
2023-05-14 16:42   ` Cédric Le Goater
2023-05-15  7:56     ` Avihai Horon
2023-05-01 14:01 ` [PATCH 3/8] migration: Add precopy initial data loaded ACK functionality Avihai Horon
2023-05-02 22:56   ` Peter Xu
2023-05-03 15:36     ` Avihai Horon
2023-05-10  8:54   ` Juan Quintela
2023-05-10 15:52     ` Avihai Horon
2023-05-10 15:59       ` Juan Quintela
2023-05-01 14:01 ` [PATCH 4/8] migration: Enable precopy initial data capability Avihai Horon
2023-05-10  8:55   ` Juan Quintela
2023-05-01 14:01 ` [PATCH 5/8] tests: Add migration precopy initial data capability test Avihai Horon
2023-05-10  8:55   ` Juan Quintela
2023-05-01 14:01 ` [PATCH 6/8] vfio/migration: Refactor vfio_save_block() to return saved data size Avihai Horon
2023-05-10  9:00   ` Juan Quintela
2023-05-01 14:01 ` [PATCH 7/8] vfio/migration: Add VFIO migration pre-copy support Avihai Horon
2023-05-01 14:01 ` [PATCH 8/8] vfio/migration: Add support for precopy initial data capability Avihai Horon
2023-05-02 22:49 ` [PATCH 0/8] migration: Add precopy initial data capability and VFIO precopy support Peter Xu
2023-05-03 15:22   ` Avihai Horon
2023-05-03 15:49     ` Peter Xu
2023-05-04 10:18       ` Avihai Horon
2023-05-04 15:50         ` Peter Xu
2023-05-07 12:54           ` Avihai Horon
2023-05-08  0:49             ` Peter Xu
2023-05-08 11:11               ` Avihai Horon
2023-05-10  9:12     ` Juan Quintela
2023-05-10 16:01       ` Avihai Horon
2023-05-10 16:41         ` Juan Quintela
2023-05-11 11:31           ` Avihai Horon
2023-05-11 13:09             ` Juan Quintela
2023-05-11 15:08               ` Avihai Horon

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=dc5af4cd-1844-5b67-2eb1-56dbee5d9dec@nvidia.com \
    --to=avihaih@nvidia.com \
    --cc=alex.williamson@redhat.com \
    --cc=armbru@redhat.com \
    --cc=clg@redhat.com \
    --cc=eblake@redhat.com \
    --cc=jgg@nvidia.com \
    --cc=joao.m.martins@oracle.com \
    --cc=kwankhede@nvidia.com \
    --cc=leobras@redhat.com \
    --cc=lvivier@redhat.com \
    --cc=maorg@nvidia.com \
    --cc=pbonzini@redhat.com \
    --cc=peterx@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=quintela@redhat.com \
    --cc=targupta@nvidia.com \
    --cc=thuth@redhat.com \
    --cc=yishaih@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).