From: "Richard W.M. Jones" <rjones(a)redhat.com&gt; --- daemon/Makefile.am | 1 + daemon/fstrim.c | 97 ++++++++++++++++++++++++++++++++++++++++ generator/generator_actions.ml | 27 ++++++++++- gobject/Makefile.inc | 6 ++- po/POTFILES | 2 + src/MAX_PROC_NR | 2 +- 6 files changed, 130 insertions(+), 5 deletions(-) create mode 100644 daemon/fstrim.c diff --git a/daemon/Makefile.am b/daemon/Makefile.am index 02c21df..9e2a633 100644 --- a/daemon/Makefile.am +++ b/daemon/Makefile.am @@ -114,6 +114,7 @@ guestfsd_SOURCES = \ fill.c \ find.c \ fsck.c \ + fstrim.c \ glob.c \ grep.c \ grub.c \ diff --git a/daemon/fstrim.c b/daemon/fstrim.c new file mode 100644 index 0000000..e2daf6a --- /dev/null +++ b/daemon/fstrim.c @@ -0,0 +1,97 @@ +/* libguestfs - the guestfsd daemon + * Copyright (C) 2012 Red Hat Inc. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +#include <config.h> + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <inttypes.h> + +#include "guestfs_protocol.h" +#include "daemon.h" +#include "actions.h" +#include "optgroups.h" + +#define MAX_ARGS 64 + +int +optgroup_fstrim_available (void) +{ + return prog_exists ("fstrim"); +} + +/* Takes optional arguments, consult optargs_bitmask. */ +int +do_fstrim (const char *path, + int64_t offset, int64_t length, int64_t minimumfreeextent) +{ + const char *argv[MAX_ARGS]; + size_t i = 0; + char offset_s[64], length_s[64], mfe_s[64]; + char *err; + int r; + + ADD_ARG (argv, i, "fstrim"); + + if ((optargs_bitmask & GUESTFS_FSTRIM_OFFSET_BITMASK)) { + if (offset < 0) { + reply_with_error ("offset < 0"); + return -1; + } + + snprintf (offset_s, sizeof offset_s, "%" PRIi64, offset); + ADD_ARG (argv, i, "-o"); + ADD_ARG (argv, i, offset_s); + } + + if ((optargs_bitmask & GUESTFS_FSTRIM_LENGTH_BITMASK)) { + if (length <= 0) { + reply_with_error ("length <= 0"); + return -1; + } + + snprintf (length_s, sizeof length_s, "%" PRIi64, length); + ADD_ARG (argv, i, "-l"); + ADD_ARG (argv, i, length_s); + } + + if ((optargs_bitmask & GUESTFS_FSTRIM_MINIMUMFREEEXTENT_BITMASK)) { + if (minimumfreeextent <= 0) { + reply_with_error ("minimumfreeextent <= 0"); + return -1; + } + + snprintf (mfe_s, sizeof mfe_s, "%" PRIi64, minimumfreeextent); + ADD_ARG (argv, i, "-m"); + ADD_ARG (argv, i, mfe_s); + } + + ADD_ARG (argv, i, path); + ADD_ARG (argv, i, NULL); + + r = commandv (NULL, &err, argv); + if (r == -1) { + reply_with_error ("%s", err); + free (err); + return -1; + } + free (err); + + return 0; +} diff --git a/generator/generator_actions.ml b/generator/generator_actions.ml index cfbce00..120278a 100644 --- a/generator/generator_actions.ml +++ b/generator/generator_actions.ml @@ -6821,8 +6821,9 @@ The filesystem must be mounted read-write. The filesystem contents are not affected, but any free space in the filesystem is freed. -In future (but not currently) these zeroed blocks will be -\"sparsified\" - that is, given back to the host."); +Free space is not \"trimmed\". You may want to call +C<guestfs_fstrim> either as an alternative to this, +or after calling this, depending on your requirements."); ("lvcreate_free", (RErr, [String "logvol"; String "volgroup"; Int "percent"], []), 312, [Optional "lvm2"], [InitEmpty, Always, TestOutputList ( @@ -7259,6 +7260,28 @@ a later version of the filesystem, or having incompatible features. See also C<guestfs_available>, L<guestfs(3)/AVAILABILITY>."); + ("fstrim", (RErr, [Pathname "mountpoint"], [OInt64 "offset"; OInt64 "length"; OInt64 "minimumfreeextent"]), 334, [Optional "fstrim"], + [], + "trim free space in a filesystem", + "\ +Trim the free space in the filesystem mounted on C<mountpoint>. +The filesystem must be mounted read-write. + +The filesystem contents are not affected, but any free space +in the filesystem is \"trimmed\", that is, given back to the host +device, thus making disk images more sparse, allowing unused space +in qcow2 files to be reused, etc. + +This operation requires support in libguestfs, the mounted +filesystem, the host filesystem, qemu and the host kernel. +If this support isn't present it may give an error or even +appear to run but do nothing. + +See also C<guestfs_zero_free_space>. That is a slightly +different operation that turns free space in the filesystem +into zeroes. It is valid to call C<guestfs_fstrim> either +instead of, or after calling C<guestfs_zero_free_space>."); + ] let all_functions = non_daemon_functions @ daemon_functions diff --git a/gobject/Makefile.inc b/gobject/Makefile.inc index d4db407..056c5f5 100644 --- a/gobject/Makefile.inc +++ b/gobject/Makefile.inc @@ -61,7 +61,8 @@ guestfs_gobject_headers= \ include/guestfs-gobject/optargs-ntfsclone_out.h \ include/guestfs-gobject/optargs-mkfs_btrfs.h \ include/guestfs-gobject/optargs-set_e2attrs.h \ - include/guestfs-gobject/optargs-btrfs_fsck.h + include/guestfs-gobject/optargs-btrfs_fsck.h \ + include/guestfs-gobject/optargs-fstrim.h guestfs_gobject_sources= \ src/session.c \ @@ -104,4 +105,5 @@ guestfs_gobject_sources= \ src/optargs-ntfsclone_out.c \ src/optargs-mkfs_btrfs.c \ src/optargs-set_e2attrs.c \ - src/optargs-btrfs_fsck.c + src/optargs-btrfs_fsck.c \ + src/optargs-fstrim.c diff --git a/po/POTFILES b/po/POTFILES index 9d3282e..747b341 100644 --- a/po/POTFILES +++ b/po/POTFILES @@ -34,6 +34,7 @@ daemon/fill.c daemon/find.c daemon/findfs.c daemon/fsck.c +daemon/fstrim.c daemon/glob.c daemon/grep.c daemon/grub.c @@ -142,6 +143,7 @@ gobject/src/optargs-copy_device_to_file.c gobject/src/optargs-copy_file_to_device.c gobject/src/optargs-copy_file_to_file.c gobject/src/optargs-e2fsck.c +gobject/src/optargs-fstrim.c gobject/src/optargs-inspect_get_icon.c gobject/src/optargs-md_create.c gobject/src/optargs-mkfs_btrfs.c diff --git a/src/MAX_PROC_NR b/src/MAX_PROC_NR index 55bd0ac..0ae9d1e 100644 --- a/src/MAX_PROC_NR +++ b/src/MAX_PROC_NR @@ -1 +1 @@ -333 +334 -- 1.7.10.1

From: "Richard W.M. Jones" <rjones(a)redhat.com&gt; QEMU 1.0 was released at the end of 2011. Remove all the cruft about detecting broken -machine type which was only required for QEMU 0.15. This also reverts commit 30ecbf3ec2ada68f7e125a180553e31b069033b7. Even on ARM you can pass -machine accel=kvm:tcg and qemu does the right thing, so I'm not sure why we wanted to disable that. --- README | 2 +- configure.ac | 68 ++++++++++++++++++---------------------------------------- src/launch.c | 27 +++++------------------ 3 files changed, 27 insertions(+), 70 deletions(-) diff --git a/README b/README index 2b133c6..4dd2075 100644 --- a/README +++ b/README @@ -46,7 +46,7 @@ For basic functionality and the C tools: - look at appliance/packagelist.in and install as many of the packages that apply to your distro as possible -- recent QEMU >= 0.13 (0.14 or later is better) with virtio-serial support +- QEMU >= 1.0. - kernel >= 2.6.34 with virtio-serial support enabled. diff --git a/configure.ac b/configure.ac index 69da9a5..ca0f9f1 100644 --- a/configure.ac +++ b/configure.ac @@ -568,55 +568,29 @@ working. ]) fi - AS_IF([test "x$QEMU_OPTIONS" = "x"],[ - dnl qemu 0.15 was released with broken support for '-machine', - dnl requiring you to add the machine type: '-machine pc,[...]'. - dnl The problem is that 'pc' is only applicable for PC-like - dnl hardware, so we cannot do this as a general solution. Since - dnl qemu 0.15, this problem has been fixed so now the default - dnl machine type is chosen (qemu commit 2645c6dcaf6ea2a51a). - dnl - dnl We need to work out if this qemu is the broken version, so we - dnl can add 'pc' just for this broken version. - dnl - dnl Note that old qemu didn't support the '-machine' option at all. - dnl - dnl We use the -kernel option for testing this, because this option - dnl is processed very late, after qemu has set up the machine. - AC_MSG_CHECKING([for broken '-machine accel=tcg' option in $QEMU]) - LC_ALL=C $QEMU -nographic -machine accel=tcg -kernel /NO_SUCH_FILE \ - > config1.tmp 2>&1 - LC_ALL=C $QEMU -nographic -machine pc,accel=tcg -kernel /NO_SUCH_FILE \ - > config2.tmp 2>&1 - if cmp -s config1.tmp config2.tmp; then - AC_MSG_RESULT([no]) - else - AC_MSG_RESULT([yes]) - AC_DEFINE([QEMU_MACHINE_TYPE_IS_BROKEN],[1],[qemu -machine accel=tcg option is broken (in qemu 0.15 only)]) - fi - rm config1.tmp config2.tmp - - dnl See if the '-machine [pc,]accel=tcg' option is required in - dnl order to run the virtio-serial test below. This happens when - dnl we run qemu-kvm inside a VM without forcing TCG: - dnl - dnl Could not access KVM kernel module: No such file or directory - dnl failed to initialize KVM: No such file or directory - dnl No accelerator found! - AC_MSG_CHECKING([if -machine @<:@pc,@:>@accel=tcg option is required to test virtio-serial feature]) - if $QEMU -nographic -device \? >/dev/null 2>&1; then - : - elif $QEMU -machine accel=tcg -nographic -device \? >/dev/null 2>&1; then - QEMU_OPTIONS_FOR_TEST="-machine accel=tcg" - elif $QEMU -machine pc,accel=tcg -nographic -device \? >/dev/null 2>&1; then - QEMU_OPTIONS_FOR_TEST="-machine pc,accel=tcg" - # else nothing ... it'll fail below. - fi - AC_MSG_RESULT([$QEMU_OPTIONS_FOR_TEST]) - ]) + AC_MSG_CHECKING([that $QEMU -version works]) + if $QEMU -version >&AS_MESSAGE_LOG_FD 2>&1; then + AC_MSG_RESULT([yes]) + else + AC_MSG_RESULT([no]) + AC_MSG_FAILURE( +[$QEMU -version: command failed. + +This could be a very old version of qemu, or qemu might not be +working. +]) + fi + + AC_MSG_CHECKING([for $QEMU version >= 1]) + if $QEMU -version | grep -sq 'version @<:@1-@:>@'; then + AC_MSG_RESULT([yes]) + else + AC_MSG_RESULT([no]) + AC_MSG_FAILURE([$QEMU version must be >= 1.0.]) + fi AC_MSG_CHECKING([for virtio-serial support in $QEMU]) - if $QEMU $QEMU_OPTIONS $QEMU_OPTIONS_FOR_TEST -nographic -device \? 2>&1 | grep -sq virtio-serial; then + if $QEMU $QEMU_OPTIONS -machine accel=kvm:tcg -device \? 2>&1 | grep -sq virtio-serial; then AC_MSG_RESULT([yes]) else AC_MSG_RESULT([no]) diff --git a/src/launch.c b/src/launch.c index 7bc913b..e175dd9 100644 --- a/src/launch.c +++ b/src/launch.c @@ -678,29 +678,13 @@ launch_appliance (guestfs_h *g) add_cmdline_shell_unquoted (g, QEMU_OPTIONS); } - /* The #if on the next line should really be "architectures for - * which KVM is commonly available. - */ -#if defined(__i386__) || defined(__x86_64__) /* 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, "-machine")) { add_cmdline (g, "-machine"); -#if QEMU_MACHINE_TYPE_IS_BROKEN - /* Workaround for qemu 0.15: We have to add the '[type=]pc' - * since there is no default. This is not a permanent solution - * because this only works on PC-like hardware. Other platforms - * like ppc would need a different machine type. - * - * This bug is fixed in qemu commit 2645c6dcaf6ea2a51a, and was - * not a problem in qemu < 0.15. - */ - add_cmdline (g, "pc,accel=kvm:tcg"); -#else add_cmdline (g, "accel=kvm:tcg"); -#endif } else { /* qemu sometimes needs this option to enable hardware * virtualization, but some versions of 'qemu-kvm' will use KVM @@ -718,7 +702,6 @@ launch_appliance (guestfs_h *g) is_openable (g, "/dev/kvm", O_RDWR|O_CLOEXEC)) add_cmdline (g, "-enable-kvm"); } -#endif /* i386 or x86-64 */ if (g->smp > 1) { snprintf (buf, sizeof buf, "%d", g->smp); @@ -1339,11 +1322,11 @@ test_qemu (guestfs_h *g) snprintf (cmd, sizeof cmd, "LC_ALL=C '%s' -nographic -help", g->qemu); - /* qemu -help should always work (qemu -version OTOH wasn't - * supported by qemu 0.9). If this command doesn't work then it - * probably indicates that the qemu binary is missing. + /* If this command doesn't work then it probably indicates that the + * qemu binary is missing. */ if (test_qemu_cmd (g, cmd, &g->qemu_help) == -1) { + qemu_error: error (g, _("command failed: %s\n\nIf qemu is located on a non-standard path, try setting the LIBGUESTFS_QEMU\nenvironment variable. There may also be errors printed above."), cmd); return -1; @@ -1352,8 +1335,8 @@ test_qemu (guestfs_h *g) snprintf (cmd, sizeof cmd, "LC_ALL=C '%s' -nographic -version 2>/dev/null", g->qemu); - /* Intentionally ignore errors from qemu -version. */ - ignore_value (test_qemu_cmd (g, cmd, &g->qemu_version)); + if (test_qemu_cmd (g, cmd, &g->qemu_version) == -1) + goto qemu_error; return 0; } -- 1.7.10.1

From: "Richard W.M. Jones" <rjones(a)redhat.com&gt; This requires febootstrap >= 3.15. --- README | 2 +- src/guestfs-internal.h | 3 ++ src/launch.c | 141 +++++++++++++++++++++++++++++++++++++++--------- 3 files changed, 121 insertions(+), 25 deletions(-) diff --git a/README b/README index 4dd2075..a17863f 100644 --- a/README +++ b/README @@ -55,7 +55,7 @@ For basic functionality and the C tools: to make complex changes to the ./configure command line to get it to work if you don't have virtio) -- febootstrap >= 3.3 (it is best to use the latest version) +- febootstrap >= 3.15 Notes: (1) febootstrap 2.x WILL NOT WORK (2) febootstrap 3.x is distro-independent, and is required on diff --git a/src/guestfs-internal.h b/src/guestfs-internal.h index fb4868c..0cdc8f2 100644 --- a/src/guestfs-internal.h +++ b/src/guestfs-internal.h @@ -257,6 +257,9 @@ struct guestfs_h int ml_read_only; /* If mounted read-only. */ int ml_debug_calls; /* Extra debug info on each FUSE call. */ #endif + + /* Used by src/launch.c:qemu_supports_virtio_scsi */ + int virtio_scsi; }; /* Per-filesystem data stored for inspect_os. */ diff --git a/src/launch.c b/src/launch.c index 3db2443..5aed29f 100644 --- a/src/launch.c +++ b/src/launch.c @@ -78,7 +78,10 @@ static int64_t timeval_diff (const struct timeval *x, const struct timeval *y); static void print_qemu_command_line (guestfs_h *g, char **argv); static int connect_unix_socket (guestfs_h *g, const char *sock); static int qemu_supports (guestfs_h *g, const char *option); -static char *qemu_drive_param (guestfs_h *g, const struct drive *drv); +static int qemu_supports_device (guestfs_h *g, const char *device_name); +static int qemu_supports_virtio_scsi (guestfs_h *g); +static char *qemu_drive_param (guestfs_h *g, const struct drive *drv, size_t index); +static char *drive_name (size_t index, char *ret); #if 0 static int qemu_supports_re (guestfs_h *g, const pcre *option_regex); @@ -248,7 +251,7 @@ guestfs__debug_drives (guestfs_h *g) ret = safe_malloc (g, sizeof (char *) * (count + 1)); for (i = 0, drv = g->drives; drv; i++, drv = drv->next) - ret[i] = qemu_drive_param (g, drv); + ret[i] = qemu_drive_param (g, drv, i); ret[count] = NULL; @@ -620,6 +623,7 @@ launch_appliance (guestfs_h *g) if (r == 0) { /* Child (qemu). */ char buf[256]; + int virtio_scsi = qemu_supports_virtio_scsi (g); /* Set up the full command line. Do this in the subprocess so we * don't need to worry about cleaning up. @@ -660,15 +664,60 @@ launch_appliance (guestfs_h *g) /* Add drives */ struct drive *drv = g->drives; + size_t drv_index = 0; + + if (virtio_scsi) { + /* Create the virtio-scsi bus. */ + add_cmdline (g, "-device"); + add_cmdline (g, "virtio-scsi-pci,id=scsi"); + } + while (drv != NULL) { /* Construct the final -drive parameter. */ - char *buf = qemu_drive_param (g, drv); + char *buf = qemu_drive_param (g, drv, drv_index); add_cmdline (g, "-drive"); add_cmdline (g, buf); free (buf); + if (qemu_supports_virtio_scsi (g)) { + char buf2[64]; + snprintf (buf2, sizeof buf2, "scsi-hd,drive=hd%zu", drv_index); + add_cmdline (g, "-device"); + add_cmdline (g, buf2); + } + drv = drv->next; + drv_index++; + } + + char appliance_root[64] = ""; + + /* Add the ext2 appliance drive (after all the drives). */ + if (appliance) { + const char *cachemode = ""; + if (qemu_supports (g, "cache=")) { + if (qemu_supports (g, "unsafe")) + cachemode = ",cache=unsafe"; + else if (qemu_supports (g, "writeback")) + cachemode = ",cache=writeback"; + } + + char buf2[PATH_MAX + 64]; + int virtio_scsi = qemu_supports_virtio_scsi (g); + add_cmdline (g, "-drive"); + snprintf (buf2, sizeof buf2, "file=%s,snapshot=on,id=appliance,if=%s%s", + appliance, virtio_scsi ? "none" : "virtio", cachemode); + add_cmdline (g, buf2); + + if (virtio_scsi) { + add_cmdline (g, "-device"); + add_cmdline (g, "scsi-hd,drive=appliance"); + } + + snprintf (appliance_root, sizeof appliance_root, "root=/dev/%cd", + virtio_scsi ? 's' : 'v'); + drive_name (drv_index, &appliance_root[12]); } if (STRNEQ (QEMU_OPTIONS, "")) { @@ -792,10 +841,12 @@ launch_appliance (guestfs_h *g) /* Linux kernel command line. */ snprintf (buf, sizeof buf, LINUX_CMDLINE + "%s " /* (root) */ "%s " /* (selinux) */ "%s " /* (verbose) */ "TERM=%s " /* (TERM environment variable) */ "%s", /* (append) */ + appliance_root, g->selinux ? "selinux=1 enforcing=0" : "selinux=0", g->verbose ? "guestfs_verbose=1" : "", getenv ("TERM") ? : "linux", @@ -808,23 +859,6 @@ launch_appliance (guestfs_h *g) add_cmdline (g, "-append"); add_cmdline (g, buf); - /* Add the ext2 appliance drive (last of all). */ - if (appliance) { - const char *cachemode = ""; - if (qemu_supports (g, "cache=")) { - if (qemu_supports (g, "unsafe")) - cachemode = ",cache=unsafe"; - else if (qemu_supports (g, "writeback")) - cachemode = ",cache=writeback"; - } - - char buf2[PATH_MAX + 64]; - add_cmdline (g, "-drive"); - snprintf (buf2, sizeof buf2, "file=%s,snapshot=on,if=virtio%s", - appliance, cachemode); - add_cmdline (g, buf2); - } - /* Finish off the command line. */ incr_cmdline_size (g); g->cmdline[g->cmdline_size-1] = NULL; @@ -1432,6 +1466,20 @@ qemu_supports_re (guestfs_h *g, const pcre *option_regex) } #endif +/* Test if device is supported by qemu (currently just greps the -device ? + * output). + */ +static int +qemu_supports_device (guestfs_h *g, const char *device_name) +{ + if (!g->qemu_devices) { + if (test_qemu (g) == -1) + return -1; + } + + return strstr (g->qemu_devices, device_name) != NULL; +} + #if defined(__i386__) || defined(__x86_64__) /* Check if a file can be opened. */ static int @@ -1447,13 +1495,39 @@ is_openable (guestfs_h *g, const char *path, int flags) } #endif +/* Returns 1 = use virtio-scsi, or 0 = use virtio-blk. */ +static int +qemu_supports_virtio_scsi (guestfs_h *g) +{ + int r; + + /* g->virtio_scsi has these values: + * 0 = untested (after handle creation) + * 1 = supported + * 2 = not supported (use virtio-blk) + * 3 = test failed (use virtio-blk) + */ + if (g->virtio_scsi == 0) { + r = qemu_supports_device (g, "virtio-scsi-pci"); + if (r > 0) + g->virtio_scsi = 1; + else if (r == 0) + g->virtio_scsi = 2; + else + g->virtio_scsi = 3; + } + + return g->virtio_scsi == 1; +} + static char * -qemu_drive_param (guestfs_h *g, const struct drive *drv) +qemu_drive_param (guestfs_h *g, const struct drive *drv, size_t index) { size_t i; - size_t len = 64; + size_t len = 128; const char *p; char *r; + const char *iface; len += strlen (drv->path) * 2; /* every "," could become ",," */ if (drv->iface) @@ -1475,16 +1549,35 @@ qemu_drive_param (guestfs_h *g, const struct drive *drv) r[i++] = *p; } - snprintf (&r[i], len-i, "%s%s%s%s,if=%s", + if (drv->iface) + iface = drv->iface; + else if (qemu_supports_virtio_scsi (g)) + iface = "none"; /* sic */ + else + iface = "virtio"; + + snprintf (&r[i], len-i, "%s%s%s%s,id=hd%zu,if=%s", drv->readonly ? ",snapshot=on" : "", drv->use_cache_off ? ",cache=off" : "", drv->format ? ",format=" : "", drv->format ? drv->format : "", - drv->iface ? drv->iface : "virtio"); + index, + iface); return r; /* caller frees */ } +/* https://rwmj.wordpress.com/2011/01/09/how-are-linux-drives-named-beyond-d... */ +static char * +drive_name (size_t index, char *ret) +{ + if (index > 26) + ret = drive_name (index / 26, ret); + index %= 26; + *ret++ = 'a' + index; + return ret; +} + /* You had to call this function after launch in versions <= 1.0.70, * but it is now a no-op. */ -- 1.7.10.1