search for: qemu_support

This set of 10 patches removes the need for any vmchannel implementation, although we can still choose to use vmchannel if we want. In this so-called "null vmchannel" configuration, the appliance connects directly to a port on the library. The exact method is described in patch 9/10. This method still requires SLIRP (user mode networking) so it is not a panacea, because recent

These assorted patches end up with adding virtio-scsi support to libguestfs. It passes libguestfs-test-tool, but I haven't yet tried to run the full set of tests. In theory > 26 devices can be added, but it's likely that certain parts of the daemon will break if you actually try this. This of course needs to be fixed. Thanks Paolo Bonzini for invaluable help. Rich.

...d *datav, const char *arg) ADD_CMDLINE (VIRTIO_BLK ".scsi=off"); } + /* This oddly named option doesn't actually enable FIPS. It just + * causes qemu to do the right thing if FIPS is enabled in the + * kernel. So like libvirt, we pass it unconditionally. + */ + if (qemu_supports (g, data, "-enable-fips")) + ADD_CMDLINE ("-enable-fips"); + if (qemu_supports (g, data, "-nodefconfig")) ADD_CMDLINE ("-nodefconfig"); -- 1.8.3.1

...;$output" - src/proto.c: replace if (!(*xdrp) (&xdr, args)) -> if (!(*xdrp) (&xdr, args, 0)) - src/proto.c: replace if (xdrp && ret && !xdrp (&xdr, ret)) -> if (xdrp && ret && !xdrp (&xdr, ret, 0)) - src/launch-direct.c: comment out - if (qemu_supports (g, data, "-nodefaults")) ADD_CMDLINE ("-nodefaults"); [This works around a qemu bug: qemu: qemu_mutex_lock: Invalid argument when using the -nodefaults parameter] - fuse/guestunmount.c: replace execlp ("fusermount", "fusermount", "-u", mountpo...

...ame) + 64; + char buf[len]; + if (strchr (filename, ',') != NULL) { error (g, _("filename cannot contain ',' (comma) character")); return -1; @@ -846,24 +849,7 @@ guestfs__add_drive_ro_with_if (guestfs_h *g, const char *filename, return -1; } - if (qemu_supports (g, NULL) == -1) - return -1; - - /* Only SCSI and virtio drivers support readonly mode. - * This is only supported as a QEMU feature since 2010/01. - */ - int supports_ro = 0; - if ((STREQ (drive_if, "scsi") || STREQ (drive_if, "virtio")) && - qemu_suppo...

...; if (cpu_model) { ADD_CMDLINE ("-cpu"); ADD_CMDLINE (cpu_model); } - /* The qemu -machine option (added 2010-12) is a bit more sane - * since it falls back through various different acceleration - * modes, so try that first (thanks Markus Armbruster). - */ - if (qemu_supports (g, data, "-machine")) { - ADD_CMDLINE ("-machine"); - if (!force_tcg) - ADD_CMDLINE ("accel=kvm:tcg"); - else - ADD_CMDLINE ("accel=tcg"); - } else { - /* qemu sometimes needs this option to enable hardware - * virtualization, but...

More comprehensive support for virtio-scsi. Passes all the tests. Rich.

...all), or M (used in > > > > master) and/or RH7 (used in rhel-7.0). > > > > > > > > > * -nographic > > > > > > > > M + RH7 > > > > I guess that could be changed into (roughly) > > > > > > > > if (qemu_supports (g, data, "-display")) > > > > { ADD_CMDLINE ("-display"); ADD_CMDLINE ("none"); } > > > > else > > > > > > > > ADD_CMDLINE ("-nographic"); > > > > > > Yup, something like that sho...

v1 -> v2: - Rebase on top of Pino's version work. Two patches went upstream, these are the two remaining patches. Note the generation number is still inside the qemu.stat file. We could put it in the filename, I have no particular preference. Rich.

...l; then AC_MSG_RESULT([yes]) else AC_MSG_RESULT([no]) diff --git a/src/launch-direct.c b/src/launch-direct.c index 8d0d8ce..f45f582 100644 --- a/src/launch-direct.c +++ b/src/launch-direct.c @@ -349,7 +349,8 @@ launch_direct (guestfs_h *g, void *datav, const char *arg) if (qemu_supports (g, data, "-nodefaults")) ADD_CMDLINE ("-nodefaults"); - ADD_CMDLINE ("-nographic"); + ADD_CMDLINE ("-display"); + ADD_CMDLINE ("none"); #ifdef MACHINE_TYPE ADD_CMDLINE ("-M"); @@ -918,7 +919,8 @@ test_qemu (guestfs_h *g,...

...utput" - src/proto.c: replace - if (!(*xdrp) (&xdr, args)) -> if (!(*xdrp) (&xdr, args, 0)) - src/proto.c: replace - if (xdrp && ret && !xdrp (&xdr, ret)) -> if (xdrp && ret && !xdrp (&xdr, ret, 0)) - src/launch-direct.c: comment out - if (qemu_supports (g, data, "-nodefaults")) ADD_CMDLINE ("-nodefaults"); - fuse/guestunmount.c: replace - execlp ("fusermount", "fusermount", "-u", mountpoint, NULL); -> execlp ("/sbin/umount", "fusermount", mountpoint, NULL); - fuse/guestunm...

Doing qemu feature detection in the direct backend takes ~100ms because we need to run `qemu -help' and `qemu -devices ?', and each of those interacts with glibc's very slow link loader. Fixing the link loader is really hard. Instead memoize the output of those two commands. This patch series first separates all the code dealing with qemu into a separate module (src/qemu.c) and

...> master) and/or RH7 (used in rhel-7.0). > > > > > > > > > > > * -nographic > > > > > > > > > > M + RH7 > > > > > I guess that could be changed into (roughly) > > > > > > > > > > if (qemu_supports (g, data, "-display")) > > > > > { ADD_CMDLINE ("-display"); ADD_CMDLINE ("none"); } > > > > > else > > > > > > > > > > ADD_CMDLINE ("-nographic"); > > > > > > > &gt...

...cmdline (g); g->cmdline[0] = g->qemu; + /* CVE-2011-4127 mitigation: Disable SCSI ioctls on virtio-blk + * devices. The -global option must exist, but you can pass any + * strings to it so we don't need to check for the specific virtio + * feature. + */ + if (qemu_supports (g, "-global")) { + add_cmdline (g, "-global"); + add_cmdline (g, "virtio-blk-pci.scsi=off"); + } + /* Add drives */ struct drive *drv = g->drives; while (drv != NULL) { -- 1.7.6

Couple of notes: (1) You may need to 'make clean' after applying this patch. There is some missing dep so it seems the appliance isn't fully rebuilt. (2) [Comment mainly directed to Matt] This is not a candidate to be automatically added to the stable 1.2 branch. It's far too large of a change for a stable release. Since this change may be required for V2V, please clone

Tested with 'make check-release'. tests/parallel (in check-slow) failed, although it does regularly and that seems to be because of libvirt. Rich.

...h *g, void *datav, const char *arg) guestfs_int_launch_send_progress (g, 3); - if (g->verbose) - guestfs_int_print_timestamped_message (g, "begin testing qemu features"); + debug (g, "begin testing qemu features"); /* Get qemu help text and version. */ if (qemu_supports (g, data, NULL) == -1) @@ -326,8 +325,7 @@ launch_direct (guestfs_h *g, void *datav, const char *arg) } } - if (g->verbose) - guestfs_int_print_timestamped_message (g, "finished testing qemu features"); + debug (g, "finished testing qemu features"); /* Co...

...launch-direct.c @@ -62,7 +62,7 @@ struct backend_direct_data { char *qemu_devices; /* Output of qemu -device ? */ /* qemu version (0, 0 if unable to parse). */ - int qemu_version_major, qemu_version_minor; + struct version qemu_version; int virtio_scsi; /* See function qemu_supports_virtio_scsi */ @@ -420,8 +420,7 @@ launch_direct (guestfs_h *g, void *datav, const char *arg) if (qemu_supports (g, data, "-no-hpet")) { ADD_CMDLINE ("-no-hpet"); } - if (data->qemu_version_major < 1 || - (data->qemu_version_major == 1 && dat...

Thanks Richard for a fast reply. Yes, indeed, im working on a nested environment. I try to run v2v inside a VM (L1) and to create an L2 by the conversion process. And on Intel. As I wrote, it fails once in few times, mainly when there is a memory pressure on L0. Kashyap, can you please share your experience? Why should it crash during nested conversion. I'm not too familiar with libguestfs

In the kernel and qemu communities it is routine for patches that will be backported to stable branches to be posted for review. I'm proposing we do the same for libguestfs stable branches. All of the attached have been tested with 'make check-release'. Rich.