From mboxrd@z Thu Jan 1 00:00:00 1970 From: Yang Hongyang 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 21:42:50 +0800 Message-ID: <55A6635A.3050904@cn.fujitsu.com> References: <1436946351-21118-1-git-send-email-yanghy@cn.fujitsu.com> <1436946351-21118-15-git-send-email-yanghy@cn.fujitsu.com> <1436964301.32371.41.camel@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1436964301.32371.41.camel@citrix.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: Ian Campbell 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 07/15/2015 08:45 PM, Ian Campbell wrote: > 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? 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. Good idea, thank you! > >> + 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) >> { > > > . > -- Thanks, Yang.