The following patch fixes some problems on powerpc Linux machines when trying to start an x86-based VM using qemu. In the order as the hunks appear the fix the following problems: - Qemu on (my) powerpc machine for some reason takes a long time for it to create its UnixIO socket. It's like 9 seconds or so and probably more once on a busy system. So increase the timeout for waiting for that socket to appear. - VM saved images written on an x86 host have a header in little endian format. So, when checking the 32 bit int for the header version and if the version is out-of-range swap the endianess of the header and try again and only then fail. This also then works when the big endian host has written the header in big endian format and now the VM is supposed to resume on a little endina host. - Setting the CPU affinity doesn't seem to work at all on (my) PPC host. So rather than failing the start of a VM when nodeGetInfo/virNodeGetInfo fail, make the setting of the CPU affinity pass on powerpc machines so the VM at least starts. Signed-off-by: Stefan Berger <stefanb(a)linux.vnet.ibm.com&gt; --- src/qemu/qemu_driver.c | 15 +++++++++++++++ src/qemu/qemu_monitor.c | 4 ++++ src/qemu/qemu_process.c | 10 +++++++++- 3 files changed, 28 insertions(+), 1 deletion(-) Index: libvirt/src/qemu/qemu_monitor.c =================================================================== --- libvirt.orig/src/qemu/qemu_monitor.c +++ libvirt/src/qemu/qemu_monitor.c @@ -249,7 +249,11 @@ qemuMonitorOpenUnix(const char *monitor) { struct sockaddr_un addr; int monfd; +#if defined(__powerpc64__) || defined(__powerpc__) + int timeout = 30; /* In seconds */ +#else int timeout = 3; /* In seconds */ +#endif int ret, i = 0; if ((monfd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { Index: libvirt/src/qemu/qemu_driver.c =================================================================== --- libvirt.orig/src/qemu/qemu_driver.c +++ libvirt/src/qemu/qemu_driver.c @@ -43,6 +43,7 @@ #include <sys/wait.h> #include <sys/ioctl.h> #include <sys/un.h> +#include <byteswap.h> #include "qemu_driver.h" @@ -1882,6 +1883,15 @@ struct qemud_save_header { int unused[15]; }; +static inline void +bswap_header(struct qemud_save_header *hdr) { + hdr->version = bswap_32(hdr->version); + hdr->xml_len = bswap_32(hdr->xml_len); + hdr->was_running = bswap_32(hdr->was_running); + hdr->compressed = bswap_32(hdr->compressed); +} + + /* return -errno on failure, or 0 on success */ static int qemuDomainSaveHeader(int fd, const char *path, char *xml, @@ -3091,6 +3101,11 @@ qemuDomainSaveImageOpen(struct qemud_dri } if (header.version > QEMUD_SAVE_VERSION) { + /* try to convert endianess */ + bswap_header(&header); + } + + if (header.version > QEMUD_SAVE_VERSION) { qemuReportError(VIR_ERR_OPERATION_FAILED, _("image version is not supported (%d > %d)"), header.version, QEMUD_SAVE_VERSION); Index: libvirt/src/qemu/qemu_process.c =================================================================== --- libvirt.orig/src/qemu/qemu_process.c +++ libvirt/src/qemu/qemu_process.c @@ -1239,7 +1239,11 @@ qemuProcessInitCpuAffinity(virDomainObjP VIR_DEBUG0("Setting CPU affinity"); if (nodeGetInfo(NULL, &nodeinfo) < 0) +#if defined (__powerpc64__) || defined (__powerpc__) + return 0; +#else return -1; +#endif /* setaffinity fails if you set bits for CPUs which * aren't present, so we have to limit ourselves */ @@ -1295,7 +1299,11 @@ qemuProcessSetVcpuAffinites(virConnectPt int vcpu, cpumaplen, hostcpus, maxcpu; if (virNodeGetInfo(conn, &nodeinfo) != 0) { - return -1; +#if defined (__powerpc64__) || defined (__powerpc__) + return 0; +#else + return -1; +#endif } if (!def->cputune.nvcpupin)