[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [PATCH] libxl: Fix migration of HVM guests
# HG changeset patch # User Ian Campbell <ian.campbell@xxxxxxxxxx> # Date 1288014999 -3600 # Node ID 3b5c6d7181fecdf6c1043a35047632ddf9950343 # Parent d7eb978c8d5ff37e5982f9ac4799bd662abd4ee1 libxl: Fix migration of HVM guests In the default "QemuDeviceModelRecord"-style HVM tail of the migration protocol the size of the qemu save record is unknown at the receiver and therefore it must read until EOF. This is not compatible with the xl migration protocol which contains a post-migration handshake and therefore cannot close the socket on the sending end. What is required is an explicit length field for the save record, which the "RemusDeviceModelState"-style HVM tail includes. Rather than overload the "RemusDeviceModelState" name for a non-Remus use case (on off chance that they need to diverge for some reason in the future) introduce a third style called "DeviceModelRecord0002" which is identical to current "RemusDeviceModelState"-style. Hopefully the inclusion of a number here will allow easier extension in the future without needing to come up with increasingly less helpful names! Also propagate errors from xc_domain_save and libxl__domain_suspend_common to callers. Signed-off-by: Ian Campbell <ian.campbell@xxxxxxxxxx> diff -r d7eb978c8d5f -r 3b5c6d7181fe tools/libxc/xc_domain_restore.c --- a/tools/libxc/xc_domain_restore.c Mon Oct 25 13:54:30 2010 +0100 +++ b/tools/libxc/xc_domain_restore.c Mon Oct 25 14:56:39 2010 +0100 @@ -494,13 +494,18 @@ static int buffer_tail_hvm(xc_interface return -1; } - /* The normal live-migration QEMU record has no length information. + /* The legacy live-migration QEMU record has no length information. * Short of reimplementing the QEMU parser, we're forced to just read - * until EOF. Remus gets around this by sending a different signature - * which includes a length prefix */ + * until EOF. + * + * Gets around this by sending a different signatures for the new + * live-migration QEMU record and Remus which includes a length + * prefix + */ if ( !memcmp(qemusig, "QemuDeviceModelRecord", sizeof(qemusig)) ) return compat_buffer_qemu(xch, ctx, fd, buf); - else if ( !memcmp(qemusig, "RemusDeviceModelState", sizeof(qemusig)) ) + else if ( !memcmp(qemusig, "DeviceModelRecord0002", sizeof(qemusig)) || + !memcmp(qemusig, "RemusDeviceModelState", sizeof(qemusig)) ) return buffer_qemu(xch, ctx, fd, buf); qemusig[20] = '\0'; diff -r d7eb978c8d5f -r 3b5c6d7181fe tools/libxc/xg_save_restore.h --- a/tools/libxc/xg_save_restore.h Mon Oct 25 13:54:30 2010 +0100 +++ b/tools/libxc/xg_save_restore.h Mon Oct 25 14:56:39 2010 +0100 @@ -104,8 +104,9 @@ * Qemu context: * char[21] : Signature: * "QemuDeviceModelRecord" : Read Qemu save data until EOF - * "RemusDeviceModelState" : uint32_t length field followed by that many + * "DeviceModelRecord0002" : uint32_t length field followed by that many * bytes of Qemu save data + * "RemusDeviceModelState" : Currently the same as "DeviceModelRecord0002". * * PV TAIL: * diff -r d7eb978c8d5f -r 3b5c6d7181fe tools/libxl/libxl.c --- a/tools/libxl/libxl.c Mon Oct 25 13:54:30 2010 +0100 +++ b/tools/libxl/libxl.c Mon Oct 25 14:56:39 2010 +0100 @@ -686,8 +686,8 @@ int libxl_domain_suspend(libxl_ctx *ctx, int debug = info != NULL && info->flags & XL_SUSPEND_DEBUG; int rc = 0; - libxl__domain_suspend_common(ctx, domid, fd, hvm, live, debug); - if (hvm) + rc = libxl__domain_suspend_common(ctx, domid, fd, hvm, live, debug); + if (!rc && hvm) rc = libxl__domain_save_device_model(ctx, domid, fd); return rc; } diff -r d7eb978c8d5f -r 3b5c6d7181fe tools/libxl/libxl_dom.c --- a/tools/libxl/libxl_dom.c Mon Oct 25 13:54:30 2010 +0100 +++ b/tools/libxl/libxl_dom.c Mon Oct 25 14:56:39 2010 +0100 @@ -442,14 +442,17 @@ int libxl__domain_suspend_common(libxl_c callbacks.switch_qemu_logdirty = libxl__domain_suspend_common_switch_qemu_logdirty; callbacks.data = &si; - xc_domain_save(ctx->xch, fd, domid, 0, 0, flags, &callbacks, hvm); + rc = xc_domain_save(ctx->xch, fd, domid, 0, 0, flags, &callbacks, hvm); + if ( rc ) { + LIBXL__LOG_ERRNO(ctx, LIBXL__LOG_ERROR, "saving domain"); + rc = ERROR_FAIL; + } if (si.suspend_eventchn > 0) xc_suspend_evtchn_release(ctx->xch, si.xce, domid, si.suspend_eventchn); if (si.xce > 0) xc_evtchn_close(si.xce); - rc = 0; out: libxl__free_all(&gc); return rc; @@ -461,15 +464,32 @@ int libxl__domain_save_device_model(libx int fd2, c; char buf[1024]; char *filename = libxl__sprintf(&gc, "/var/lib/xen/qemu-save.%d", domid); + struct stat st; + uint32_t qemu_state_len; LIBXL__LOG(ctx, LIBXL__LOG_DEBUG, "Saving device model state to %s", filename); libxl__xs_write(&gc, XBT_NULL, libxl__sprintf(&gc, "/local/domain/0/device-model/%d/command", domid), "save"); libxl__wait_for_device_model(ctx, domid, "paused", NULL, NULL); + if (stat(filename, &st) < 0) + { + LIBXL__LOG(ctx, LIBXL__LOG_ERROR, "Unable to stat qemu save file\n"); + return ERROR_FAIL; + } + + qemu_state_len = st.st_size; + LIBXL__LOG(ctx, LIBXL__LOG_DEBUG, "Qemu state is %d bytes\n", qemu_state_len); + c = libxl_write_exactly(ctx, fd, QEMU_SIGNATURE, strlen(QEMU_SIGNATURE), "saved-state file", "qemu signature"); if (c) return c; + + c = libxl_write_exactly(ctx, fd, &qemu_state_len, sizeof(qemu_state_len), + "saved-state file", "saved-state length"); + if (c) + return c; + fd2 = open(filename, O_RDONLY); while ((c = read(fd2, buf, sizeof(buf))) != 0) { if (c < 0) { diff -r d7eb978c8d5f -r 3b5c6d7181fe tools/libxl/libxl_internal.h --- a/tools/libxl/libxl_internal.h Mon Oct 25 13:54:30 2010 +0100 +++ b/tools/libxl/libxl_internal.h Mon Oct 25 14:56:39 2010 +0100 @@ -44,7 +44,7 @@ #define LIBXL_PV_EXTRA_MEMORY 1024 #define LIBXL_HVM_EXTRA_MEMORY 2048 #define LIBXL_MIN_DOM0_MEM (128*1024) -#define QEMU_SIGNATURE "QemuDeviceModelRecord" +#define QEMU_SIGNATURE "DeviceModelRecord0002" #define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |