From: aliguori Date: Wed, 22 Apr 2009 15:19:44 +0000 (+0000) Subject: xen: pv domain builder. (Gerd Hoffmann) X-Git-Tag: 0.10.0-0sb10~1427 X-Git-Url: http://vcs.maemo.org/git/?a=commitdiff_plain;h=9306acb50991b7c78510ff4ffe19a190f2fc89c0;p=qemu xen: pv domain builder. (Gerd Hoffmann) This adds domain building support for paravirtual domains to qemu. This allows booting xen guests directly with qemu, without Xend and the management stack. Signed-off-by: Gerd Hoffmann Signed-off-by: Anthony Liguori git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7226 c046a42c-6fe2-441c-8c8c-71466251a162 --- diff --git a/Makefile.target b/Makefile.target index 24d409e..2171587 100644 --- a/Makefile.target +++ b/Makefile.target @@ -561,7 +561,7 @@ LIBS += $(CONFIG_BLUEZ_LIBS) endif # xen backend driver support -XEN_OBJS := xen_machine_pv.o xen_backend.o xen_devconfig.o +XEN_OBJS := xen_machine_pv.o xen_backend.o xen_devconfig.o xen_domainbuild.o XEN_OBJS += xen_console.o xenfb.o xen_disk.o xen_nic.o ifeq ($(CONFIG_XEN), yes) OBJS += $(XEN_OBJS) diff --git a/configure b/configure index ec761ff..b7b7b01 100755 --- a/configure +++ b/configure @@ -1634,7 +1634,7 @@ if test "$bluez" = "yes" ; then echo "#define CONFIG_BLUEZ 1" >> $config_h fi if test "$xen" = "yes" ; then - echo "XEN_LIBS=-lxenstore -lxenctrl" >> $config_mak + echo "XEN_LIBS=-lxenstore -lxenctrl -lxenguest" >> $config_mak fi if test "$aio" = "yes" ; then echo "#define CONFIG_AIO 1" >> $config_h diff --git a/hw/xen_backend.h b/hw/xen_backend.h index 7f1804e..4dbfdb4 100644 --- a/hw/xen_backend.h +++ b/hw/xen_backend.h @@ -100,5 +100,8 @@ void xen_init_display(int domid); void xen_config_cleanup(void); int xen_config_dev_blk(DriveInfo *disk); int xen_config_dev_nic(NICInfo *nic); +int xen_config_dev_vfb(int vdev, const char *type); +int xen_config_dev_vkbd(int vdev); +int xen_config_dev_console(int vdev); #endif /* QEMU_HW_XEN_BACKEND_H */ diff --git a/hw/xen_devconfig.c b/hw/xen_devconfig.c index 4420561..4121188 100644 --- a/hw/xen_devconfig.c +++ b/hw/xen_devconfig.c @@ -140,3 +140,32 @@ int xen_config_dev_nic(NICInfo *nic) /* common stuff */ return xen_config_dev_all(fe, be); } + +int xen_config_dev_vfb(int vdev, const char *type) +{ + char fe[256], be[256]; + + xen_config_dev_dirs("vfb", "vfb", vdev, fe, be, sizeof(fe)); + + /* backend */ + xenstore_write_str(be, "type", type); + + /* common stuff */ + return xen_config_dev_all(fe, be); +} + +int xen_config_dev_vkbd(int vdev) +{ + char fe[256], be[256]; + + xen_config_dev_dirs("vkbd", "vkbd", vdev, fe, be, sizeof(fe)); + return xen_config_dev_all(fe, be); +} + +int xen_config_dev_console(int vdev) +{ + char fe[256], be[256]; + + xen_config_dev_dirs("console", "console", vdev, fe, be, sizeof(fe)); + return xen_config_dev_all(fe, be); +} diff --git a/hw/xen_domainbuild.c b/hw/xen_domainbuild.c new file mode 100644 index 0000000..21710eb --- /dev/null +++ b/hw/xen_domainbuild.c @@ -0,0 +1,294 @@ +#include +#include "xen_backend.h" +#include "xen_domainbuild.h" +#include "sysemu.h" +#include "qemu-timer.h" + +#include + +static int xenstore_domain_mkdir(char *path) +{ + struct xs_permissions perms_ro[] = {{ + .id = 0, /* set owner: dom0 */ + },{ + .id = xen_domid, + .perms = XS_PERM_READ, + }}; + struct xs_permissions perms_rw[] = {{ + .id = 0, /* set owner: dom0 */ + },{ + .id = xen_domid, + .perms = XS_PERM_READ | XS_PERM_WRITE, + }}; + const char *writable[] = { "device", "control", "error", NULL }; + char subpath[256]; + int i; + + if (!xs_mkdir(xenstore, 0, path)) { + fprintf(stderr, "%s: xs_mkdir %s: failed\n", __FUNCTION__, path); + return -1; + } + if (!xs_set_permissions(xenstore, 0, path, perms_ro, 2)) { + fprintf(stderr, "%s: xs_set_permissions failed\n", __FUNCTION__); + return -1; + } + + for (i = 0; writable[i]; i++) { + snprintf(subpath, sizeof(subpath), "%s/%s", path, writable[i]); + if (!xs_mkdir(xenstore, 0, subpath)) { + fprintf(stderr, "%s: xs_mkdir %s: failed\n", __FUNCTION__, subpath); + return -1; + } + if (!xs_set_permissions(xenstore, 0, subpath, perms_rw, 2)) { + fprintf(stderr, "%s: xs_set_permissions failed\n", __FUNCTION__); + return -1; + } + } + return 0; +} + +int xenstore_domain_init1(const char *kernel, const char *ramdisk, + const char *cmdline) +{ + char *dom, uuid_string[42], vm[256], path[256]; + int i; + + snprintf(uuid_string, sizeof(uuid_string), UUID_FMT, + qemu_uuid[0], qemu_uuid[1], qemu_uuid[2], qemu_uuid[3], + qemu_uuid[4], qemu_uuid[5], qemu_uuid[6], qemu_uuid[7], + qemu_uuid[8], qemu_uuid[9], qemu_uuid[10], qemu_uuid[11], + qemu_uuid[12], qemu_uuid[13], qemu_uuid[14], qemu_uuid[15]); + dom = xs_get_domain_path(xenstore, xen_domid); + snprintf(vm, sizeof(vm), "/vm/%s", uuid_string); + + xenstore_domain_mkdir(dom); + + xenstore_write_str(vm, "image/ostype", "linux"); + if (kernel) + xenstore_write_str(vm, "image/kernel", kernel); + if (ramdisk) + xenstore_write_str(vm, "image/ramdisk", ramdisk); + if (cmdline) + xenstore_write_str(vm, "image/cmdline", cmdline); + + /* name + id */ + xenstore_write_str(vm, "name", qemu_name ? qemu_name : "no-name"); + xenstore_write_str(vm, "uuid", uuid_string); + xenstore_write_str(dom, "name", qemu_name ? qemu_name : "no-name"); + xenstore_write_int(dom, "domid", xen_domid); + xenstore_write_str(dom, "vm", vm); + + /* memory */ + xenstore_write_int(dom, "memory/target", ram_size >> 10); // kB + xenstore_write_int(vm, "memory", ram_size >> 20); // MB + xenstore_write_int(vm, "maxmem", ram_size >> 20); // MB + + /* cpus */ + for (i = 0; i < smp_cpus; i++) { + snprintf(path, sizeof(path), "cpu/%d/availability",i); + xenstore_write_str(dom, path, "online"); + } + xenstore_write_int(vm, "vcpu_avail", smp_cpus); + xenstore_write_int(vm, "vcpus", smp_cpus); + + /* vnc password */ + xenstore_write_str(vm, "vncpassword", "" /* FIXME */); + + free(dom); + return 0; +} + +int xenstore_domain_init2(int xenstore_port, int xenstore_mfn, + int console_port, int console_mfn) +{ + char *dom; + + dom = xs_get_domain_path(xenstore, xen_domid); + + /* signal new domain */ + xs_introduce_domain(xenstore, + xen_domid, + xenstore_mfn, + xenstore_port); + + /* xenstore */ + xenstore_write_int(dom, "store/ring-ref", xenstore_mfn); + xenstore_write_int(dom, "store/port", xenstore_port); + + /* console */ + xenstore_write_str(dom, "console/type", "ioemu"); + xenstore_write_int(dom, "console/limit", 128 * 1024); + xenstore_write_int(dom, "console/ring-ref", console_mfn); + xenstore_write_int(dom, "console/port", console_port); + xen_config_dev_console(0); + + free(dom); + return 0; +} + +/* ------------------------------------------------------------- */ + +static QEMUTimer *xen_poll; + +/* check domain state once per second */ +static void xen_domain_poll(void *opaque) +{ + struct xc_dominfo info; + int rc; + + rc = xc_domain_getinfo(xen_xc, xen_domid, 1, &info); + if ((1 != rc) || (info.domid != xen_domid)) { + qemu_log("xen: domain %d is gone\n", xen_domid); + goto quit; + } + if (info.dying) { + qemu_log("xen: domain %d is dying (%s%s)\n", xen_domid, + info.crashed ? "crashed" : "", + info.shutdown ? "shutdown" : ""); + goto quit; + } + + qemu_mod_timer(xen_poll, qemu_get_clock(rt_clock) + 1000); + return; + +quit: + qemu_system_shutdown_request(); + return; +} + +static void xen_domain_watcher(void) +{ + int qemu_running = 1; + int fd[2], i, n, rc; + char byte; + + pipe(fd); + if (fork() != 0) + return; /* not child */ + + /* close all file handles, except stdio/out/err, + * our watch pipe and the xen interface handle */ + n = getdtablesize(); + for (i = 3; i < n; i++) { + if (i == fd[0]) + continue; + if (i == xen_xc) + continue; + close(i); + } + + /* ignore term signals */ + signal(SIGINT, SIG_IGN); + signal(SIGTERM, SIG_IGN); + + /* wait for qemu exiting */ + while (qemu_running) { + rc = read(fd[0], &byte, 1); + switch (rc) { + case -1: + if (EINTR == errno) + continue; + qemu_log("%s: Huh? read error: %s\n", __FUNCTION__, strerror(errno)); + qemu_running = 0; + break; + case 0: + /* EOF -> qemu exited */ + qemu_running = 0; + break; + default: + qemu_log("%s: Huh? data on the watch pipe?\n", __FUNCTION__); + break; + } + } + + /* cleanup */ + qemu_log("%s: destroy domain %d\n", __FUNCTION__, xen_domid); + xc_domain_destroy(xen_xc, xen_domid); + _exit(0); +} + +/* normal cleanup */ +static void xen_domain_cleanup(void) +{ + char *dom; + + dom = xs_get_domain_path(xenstore, xen_domid); + if (dom) { + xs_rm(xenstore, 0, dom); + free(dom); + } + xs_release_domain(xenstore, xen_domid); +} + +int xen_domain_build_pv(const char *kernel, const char *ramdisk, + const char *cmdline) +{ + uint32_t ssidref = 0; + uint32_t flags = 0; + xen_domain_handle_t uuid; + unsigned int xenstore_port = 0, console_port = 0; + unsigned long xenstore_mfn = 0, console_mfn = 0; + int rc; + + memcpy(uuid, qemu_uuid, sizeof(uuid)); + rc = xc_domain_create(xen_xc, ssidref, uuid, flags, &xen_domid); + if (rc < 0) { + fprintf(stderr, "xen: xc_domain_create() failed\n"); + goto err; + } + qemu_log("xen: created domain %d\n", xen_domid); + atexit(xen_domain_cleanup); + xen_domain_watcher(); + + xenstore_domain_init1(kernel, ramdisk, cmdline); + + rc = xc_domain_max_vcpus(xen_xc, xen_domid, smp_cpus); + if (rc < 0) { + fprintf(stderr, "xen: xc_domain_max_vcpus() failed\n"); + goto err; + } + +#if 0 + rc = xc_domain_setcpuweight(xen_xc, xen_domid, 256); + if (rc < 0) { + fprintf(stderr, "xen: xc_domain_setcpuweight() failed\n"); + goto err; + } +#endif + + rc = xc_domain_setmaxmem(xen_xc, xen_domid, ram_size >> 10); + if (rc < 0) { + fprintf(stderr, "xen: xc_domain_setmaxmem() failed\n"); + goto err; + } + + xenstore_port = xc_evtchn_alloc_unbound(xen_xc, xen_domid, 0); + console_port = xc_evtchn_alloc_unbound(xen_xc, xen_domid, 0); + + rc = xc_linux_build(xen_xc, xen_domid, ram_size >> 20, + kernel, ramdisk, cmdline, + 0, flags, + xenstore_port, &xenstore_mfn, + console_port, &console_mfn); + if (rc < 0) { + fprintf(stderr, "xen: xc_linux_build() failed\n"); + goto err; + } + + xenstore_domain_init2(xenstore_port, xenstore_mfn, + console_port, console_mfn); + + qemu_log("xen: unpausing domain %d\n", xen_domid); + rc = xc_domain_unpause(xen_xc, xen_domid); + if (rc < 0) { + fprintf(stderr, "xen: xc_domain_unpause() failed\n"); + goto err; + } + + xen_poll = qemu_new_timer(rt_clock, xen_domain_poll, NULL); + qemu_mod_timer(xen_poll, qemu_get_clock(rt_clock) + 1000); + return 0; + +err: + return -1; +} diff --git a/hw/xen_domainbuild.h b/hw/xen_domainbuild.h new file mode 100644 index 0000000..dea0121 --- /dev/null +++ b/hw/xen_domainbuild.h @@ -0,0 +1,13 @@ +#ifndef QEMU_HW_XEN_DOMAINBUILD_H +#define QEMU_HW_XEN_DOMAINBUILD_H 1 + +#include "xen_common.h" + +int xenstore_domain_init1(const char *kernel, const char *ramdisk, + const char *cmdline); +int xenstore_domain_init2(int xenstore_port, int xenstore_mfn, + int console_port, int console_mfn); +int xen_domain_build_pv(const char *kernel, const char *ramdisk, + const char *cmdline); + +#endif /* QEMU_HW_XEN_DOMAINBUILD_H */ diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c index 154cf71..56b01aa 100644 --- a/hw/xen_machine_pv.c +++ b/hw/xen_machine_pv.c @@ -27,6 +27,7 @@ #include "sysemu.h" #include "boards.h" #include "xen_backend.h" +#include "xen_domainbuild.h" uint32_t xen_domid; enum xen_mode xen_mode = XEN_EMULATE; @@ -57,6 +58,24 @@ static void xen_init_pv(ram_addr_t ram_size, int vga_ram_size, fprintf(stderr, "%s: xen backend core setup failed\n", __FUNCTION__); exit(1); } + + switch (xen_mode) { + case XEN_ATTACH: + /* nothing to do, xend handles everything */ + break; + case XEN_CREATE: + if (xen_domain_build_pv(kernel_filename, initrd_filename, + kernel_cmdline) < 0) { + fprintf(stderr, "xen pv domain creation failed\n"); + exit(1); + } + break; + case XEN_EMULATE: + fprintf(stderr, "xen emulation not implemented (yet)\n"); + exit(1); + break; + } + xen_be_register("console", &xen_console_ops); xen_be_register("vkbd", &xen_kbdmouse_ops); xen_be_register("vfb", &xen_framebuffer_ops);