From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Campbell Subject: Re: [PATCH v4 --for 4.6 COLOPre 14/25] tools/libxl: introduce libxl__domain_restore_device_model to load qemu state Date: Wed, 15 Jul 2015 13:45:01 +0100 Message-ID: <1436964301.32371.41.camel@citrix.com> References: <1436946351-21118-1-git-send-email-yanghy@cn.fujitsu.com> <1436946351-21118-15-git-send-email-yanghy@cn.fujitsu.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1436946351-21118-15-git-send-email-yanghy@cn.fujitsu.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Yang Hongyang Cc: wei.liu2@citrix.com, wency@cn.fujitsu.com, andrew.cooper3@citrix.com, yunhong.jiang@intel.com, eddie.dong@intel.com, xen-devel@lists.xen.org, rshriram@cs.ubc.ca, guijianfeng@cn.fujitsu.com, Anthony Perard , ian.jackson@eu.citrix.com List-Id: xen-devel@lists.xenproject.org On Wed, 2015-07-15 at 15:45 +0800, Yang Hongyang wrote: > In normal migration, the qemu state was passed to qemu as a parameter. > With COLO, Secondary vm is running. So we will do the following steps > at every checkpoint: > 1. suspend both primay vm and secondary vm "primary" > 2. sync the state > 3. resume both primary vm and secondary vm > Primary will send qemu's state in step2, and > Secondary's qemu should read it and restore the state before it > is resumed. We can not pass the state to qemu as a parameter because > Secondary QEMU already started at this point, so we introduce > libxl__domain_restore_device_model() to do it. > This API should be called before resuming secondary vm. > > Signed-off-by: Yang Hongyang > Signed-off-by: Wen Congyang > Cc: Anthony Perard > --- > tools/libxl/libxl_dom_save.c | 29 +++++++++++++++++++++++++++++ > tools/libxl/libxl_internal.h | 3 +++ > tools/libxl/libxl_qmp.c | 10 ++++++++++ > 3 files changed, 42 insertions(+) > > diff --git a/tools/libxl/libxl_dom_save.c b/tools/libxl/libxl_dom_save.c > index f89f5d4..0926b71 100644 > --- a/tools/libxl/libxl_dom_save.c > +++ b/tools/libxl/libxl_dom_save.c > @@ -675,6 +675,35 @@ out: > return ret; > } > > +int libxl__domain_restore_device_model(libxl__gc *gc, uint32_t domid) > +{ > + char *state_file; > + int rc; > + > + switch (libxl__device_model_version_running(gc, domid)) { > + case LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN_TRADITIONAL: > + /* not supported now */ > + rc = ERROR_INVAL; > + break; > + case LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN: > + /* > + * This function may be called too many times for the same gc, > + * so we use NOGC, and free the memory before return to avoid > + * OOM. > + */ It occurs to me that domid shouldn't change for the duration of a COLO run, right? Thus I think the path can be allocated once at start of day and not per iteration, and can be stored in suspend_state (or similar) and passed in here. Hence no complexity like a nested ao is needed. > + state_file = libxl__sprintf(NOGC, > + XC_DEVICE_MODEL_RESTORE_FILE".%d", > + domid); > + rc = libxl__qmp_restore(gc, domid, state_file); > + free(state_file); > + break; > + default: > + rc = ERROR_INVAL; > + } > + > + return rc; > +} > + > /* > * Local variables: > * mode: C > diff --git a/tools/libxl/libxl_internal.h b/tools/libxl/libxl_internal.h > index 0eb5f41..fb777c1 100644 > --- a/tools/libxl/libxl_internal.h > +++ b/tools/libxl/libxl_internal.h > @@ -1074,6 +1074,7 @@ _hidden int libxl__domain_rename(libxl__gc *gc, uint32_t domid, > > _hidden int libxl__toolstack_restore(uint32_t domid, const uint8_t *buf, > uint32_t size, void *data); > +_hidden int libxl__domain_restore_device_model(libxl__gc *gc, uint32_t domid); > _hidden int libxl__domain_resume_device_model(libxl__gc *gc, uint32_t domid); > > _hidden const char *libxl__userdata_path(libxl__gc *gc, uint32_t domid, > @@ -1702,6 +1703,8 @@ _hidden int libxl__qmp_stop(libxl__gc *gc, int domid); > _hidden int libxl__qmp_resume(libxl__gc *gc, int domid); > /* Save current QEMU state into fd. */ > _hidden int libxl__qmp_save(libxl__gc *gc, int domid, const char *filename); > +/* Load current QEMU state from fd. */ > +_hidden int libxl__qmp_restore(libxl__gc *gc, int domid, const char *filename); > /* Set dirty bitmap logging status */ > _hidden int libxl__qmp_set_global_dirty_log(libxl__gc *gc, int domid, bool enable); > _hidden int libxl__qmp_insert_cdrom(libxl__gc *gc, int domid, const libxl_device_disk *disk); > diff --git a/tools/libxl/libxl_qmp.c b/tools/libxl/libxl_qmp.c > index 6484f5e..080cb9f 100644 > --- a/tools/libxl/libxl_qmp.c > +++ b/tools/libxl/libxl_qmp.c > @@ -904,6 +904,16 @@ int libxl__qmp_save(libxl__gc *gc, int domid, const char *filename) > NULL, NULL); > } > > +int libxl__qmp_restore(libxl__gc *gc, int domid, const char *state_file) > +{ > + libxl__json_object *args = NULL; > + > + qmp_parameters_add_string(gc, &args, "filename", state_file); > + > + return qmp_run_command(gc, domid, "xen-load-devices-state", args, > + NULL, NULL); > +} > + > static int qmp_change(libxl__gc *gc, libxl__qmp_handler *qmp, > char *device, char *target, char *arg) > {