July 2014 - Libguestfs - libguestfs List Archives

[PATCH] make-fs: respect libguestfs' temporary dir
by Pino Toscano 28 Jul '14

28 Jul '14

Do not hardcode /tmp. --- make-fs/make-fs.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/make-fs/make-fs.c b/make-fs/make-fs.c index 1bec3e2..9c11fef 100644 --- a/make-fs/make-fs.c +++ b/make-fs/make-fs.c @@ -381,8 +381,9 @@ estimate_input (const char *input, uint64_t *estimate_rtn, char **ifmt_rtn) CLEANUP_FCLOSE FILE *fp = NULL; char line[256]; size_t len; + CLEANUP_FREE char *tmpdir = guestfs_get_tmpdir (g); - if (asprintf (&tmpfile, "/tmp/makefsXXXXXX") == -1) { + if (asprintf (&tmpfile, "%s/makefsXXXXXX", tmpdir) == -1) { perror ("asprintf"); return -1; } -- 1.9.3

2 5

[PATCH] launch: Close file descriptors after fork (RHBZ#1123007).
by Richard W.M. Jones 28 Jul '14

28 Jul '14

This refactors existing code to close file descriptors in the recovery process, and also adds code to close file descriptors between the fork() and exec() of QEMU or User-Mode Linux. The reason is to avoid leaking main process file descriptors where the main process (or other libraries in the main process) are not setting O_CLOEXEC at all or not setting it atomically. Python is a particular culprit. See also this OpenStack Nova bug report: https://bugs.launchpad.net/nova/+bug/1313477 Thanks: Qin Zhao for identifying and characterizing the problem in Nova. --- src/guestfs-internal-frontend.h | 16 +++++++++++++++- src/launch-direct.c | 17 +++++++++-------- src/launch-uml.c | 13 +++++-------- 3 files changed, 29 insertions(+), 17 deletions(-) diff --git a/src/guestfs-internal-frontend.h b/src/guestfs-internal-frontend.h index 6bf0a94..3129018 100644 --- a/src/guestfs-internal-frontend.h +++ b/src/guestfs-internal-frontend.h @@ -1,5 +1,5 @@ /* libguestfs - * Copyright (C) 2013 Red Hat Inc. + * Copyright (C) 2013-2014 Red Hat Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -161,4 +161,18 @@ extern GUESTFS_DLL_PUBLIC int guestfs___add_libvirt_dom (guestfs_h *g, virDomain # define program_name "libguestfs" #endif +/* Close all file descriptors matching the condition. */ +#define close_file_descriptors(cond) do { \ + int max_fd = sysconf (_SC_OPEN_MAX); \ + int fd; \ + if (max_fd == -1) \ + max_fd = 1024; \ + if (max_fd > 65536) \ + max_fd = 65536; /* bound the amount of work we do here */ \ + for (fd = 0; fd < max_fd; ++fd) { \ + if (cond) \ + close (fd); \ + } \ + } while (0) + #endif /* GUESTFS_INTERNAL_FRONTEND_H_ */ diff --git a/src/launch-direct.c b/src/launch-direct.c index bb73d19..104809d 100644 --- a/src/launch-direct.c +++ b/src/launch-direct.c @@ -717,6 +717,13 @@ launch_direct (guestfs_h *g, void *datav, const char *arg) goto dup_failed; close (sv[1]); + + /* Close any other file descriptors that we don't want to pass + * to qemu. This prevents file descriptors which didn't have + * O_CLOEXEC set properly from leaking into the subprocess. See + * RHBZ#1123007. + */ + close_file_descriptors (fd >= 2); } /* Dump the command line (after setting up stderr above). */ @@ -747,7 +754,7 @@ launch_direct (guestfs_h *g, void *datav, const char *arg) if (g->recovery_proc) { r = fork (); if (r == 0) { - int i, fd, max_fd; + int i; struct sigaction sa; pid_t qemu_pid = data->pid; pid_t parent_pid = getppid (); @@ -767,13 +774,7 @@ launch_direct (guestfs_h *g, void *datav, const char *arg) /* Close all other file descriptors. This ensures that we don't * hold open (eg) pipes from the parent process. */ - max_fd = sysconf (_SC_OPEN_MAX); - if (max_fd == -1) - max_fd = 1024; - if (max_fd > 65536) - max_fd = 65536; /* bound the amount of work we do here */ - for (fd = 0; fd < max_fd; ++fd) - close (fd); + close_file_descriptors (1); /* It would be nice to be able to put this in the same process * group as qemu (ie. setpgid (0, qemu_pid)). However this is diff --git a/src/launch-uml.c b/src/launch-uml.c index 2a6ddaf..88c684b 100644 --- a/src/launch-uml.c +++ b/src/launch-uml.c @@ -333,6 +333,9 @@ launch_uml (guestfs_h *g, void *datav, const char *arg) goto dup_failed; close (csv[1]); + + /* RHBZ#1123007 */ + close_file_descriptors (fd >= 2 && fd != dsv[1]); } /* Dump the command line (after setting up stderr above). */ @@ -360,7 +363,7 @@ launch_uml (guestfs_h *g, void *datav, const char *arg) if (g->recovery_proc) { r = fork (); if (r == 0) { - int i, fd, max_fd; + int i; struct sigaction sa; pid_t vmlinux_pid = data->pid; pid_t parent_pid = getppid (); @@ -380,13 +383,7 @@ launch_uml (guestfs_h *g, void *datav, const char *arg) /* Close all other file descriptors. This ensures that we don't * hold open (eg) pipes from the parent process. */ - max_fd = sysconf (_SC_OPEN_MAX); - if (max_fd == -1) - max_fd = 1024; - if (max_fd > 65536) - max_fd = 65536; /* bound the amount of work we do here */ - for (fd = 0; fd < max_fd; ++fd) - close (fd); + close_file_descriptors (1); /* It would be nice to be able to put this in the same process * group as vmlinux (ie. setpgid (0, vmlinux_pid)). However -- 1.9.0

3 4

[PATCH] p2v: properly call va_end
by Pino Toscano 26 Jul '14

26 Jul '14

--- p2v/miniexpect.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/p2v/miniexpect.c b/p2v/miniexpect.c index ec81030..1baab5f 100644 --- a/p2v/miniexpect.c +++ b/p2v/miniexpect.c @@ -114,6 +114,7 @@ mexp_spawnl (const char *file, const char *arg, ...) new_argv = realloc (argv, sizeof (char *) * (i+1)); if (new_argv == NULL) { free (argv); + va_end (args); return NULL; } argv = new_argv; @@ -122,6 +123,7 @@ mexp_spawnl (const char *file, const char *arg, ...) h = mexp_spawnv (file, argv); free (argv); + va_end (args); return h; } -- 1.9.3

2 4

[PATCH] sparsify: Relax requirement that output device cannot be block device.
by Richard W.M. Jones 23 Jul '14

23 Jul '14

To fix RHBZ#1056290, I prevented virt-sparsify being used if the output device is a block device. I have now retested this scenario and it does work (in both copying and in-place mode), and does not delete the output device, and therefore we can relax this restriction so only char devices are banned. This is useful for oVirt which uses a qcow2 formatted block device to store virtual machines. --- sparsify/cmdline.ml | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) diff --git a/sparsify/cmdline.ml b/sparsify/cmdline.ml index a99c851..b1fb83f 100644 --- a/sparsify/cmdline.ml +++ b/sparsify/cmdline.ml @@ -167,11 +167,7 @@ read the man page virt-sparsify(1). else Sys.getcwd () // indisk in - (* Check the output is not a block or char special (RHBZ#1056290). *) - if is_block_device outdisk then - error (f_"output '%s' cannot be a block device, it must be a regular file") - outdisk; - + (* Check the output is not a char special (RHBZ#1056290). *) if is_char_device outdisk then error (f_"output '%s' cannot be a character device, it must be a regular file") outdisk; -- 1.9.0

1 0

Reminder: libguestfs preview for RHEL 7.1 is available
by Richard W.M. Jones 22 Jul '14

22 Jul '14

Just a reminder that if you're using RHEL 7.0 or CentOS 7.0, you might be interested in using the RHEL 7.1 libguestfs preview packages. The general plan is to rebase to libguestfs 1.28 when it is available. I have just updated them to: - libguestfs 1.27.22 - hivex 1.3.10 - supermin 5.1.9 To install these packages, follow the simple instructions here: https://www.redhat.com/archives/libguestfs/2014-May/msg00090.html Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into KVM guests. http://libguestfs.org/virt-v2v

1 0

IRC question: Ruby gems
by Richard W.M. Jones 20 Jul '14

20 Jul '14

18:24 < cknapp> rwmjones: is the RPM the best way to grab the guestfs gem? [Please stay in the channel after asking a question!] We currently ship the Ruby bindings as an extension in the ruby/ subdirectory of the source tarball. (Note that because the extension is mostly generated during the build, you will have to download the tarball from http://libguestfs.org/download/ which contains the generated bits; or, much less conveniently, you will have to do a full build from git). So there is no gem as such. In particular there is nothing on rubygems.org. However it should be possible to take the ruby/ subdirectory and turn it into a gem with minimal effort, since it is already in mostly the right format. In particular, see the Gem::Specification section in ruby/Rakefile. I don't know how the gem relates to the Fedora RPM. Possibly not at all. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-builder quickly builds VMs from scratch http://libguestfs.org/virt-builder.1.html

1 1

[PATCH] configure: look for "default-java" as directory for Java
by Pino Toscano 17 Jul '14

17 Jul '14

In Debian-based systems, /usr/lib/jvm/default-java is a symlink pointing to the location of the default Java version. --- configure.ac | 1 + 1 file changed, 1 insertion(+) diff --git a/configure.ac b/configure.ac index ec98e4b..932bad1 100644 --- a/configure.ac +++ b/configure.ac @@ -1250,6 +1250,7 @@ if test "x$with_java" != "xno"; then # Look for Java in some likely locations. for d in \ /usr/lib/jvm/java \ + /usr/lib/jvm/default-java \ /usr/lib/jvm/java-6-openjdk do if test -d $d && test -f $d/bin/java; then -- 1.9.3

2 1

working with mount-local
by Vasiliy Tolstov 16 Jul '14

16 Jul '14

Hi. I'm try to use guestfish inside golang (extract some files from image and create squashfs from root). Now i can't understand how can i work with mount-local/mount-local-run. Where i can specify what partition i'm export to local? Does i need firstly mount /dev/sda1 to / for example and after that run mount-local /tmp/xxx ? Second - i'm work with guestfish via writing to stdin and getting result from stdout. In case of mount-local-run does i need fully fork new binary or i can only dup stdin descriptor and work with it? -- Vasiliy Tolstov, e-mail: v.tolstov(a)selfip.ru jabber: vase(a)selfip.ru

3 9

supermin: how can I avoid downloading packages everytime when run make
by Hu Tao 16 Jul '14

16 Jul '14

Hi, When I run make in libguestfs directory, supermin tries to redownload all packages it needs everytime. This is a blocker for me because: 1. I have a slow network connection, redownloading all packages is just a waste of time even though they are not big. 2. My network connection is not stable(especially when visiting sites outside of China), sometimes it fails to download the packages in the middle of downloading, so I have to restart from the beginning. 3. I'm not expected to consume too much network traffic from within office. So what I want to know is: 1. is it possible to use cached packages that supermin already downloaded? How? 2. or better, whether supermin can use packages installed on local systems, instead of downloading from web? I read the man page of supermin and saw there is --if-newer and --use-installed, but I'm not sure how to use them. Any helps will be appreciated! Regards, Hu Tao

2 4

[PATCH] appliance: init: run ldconfig
by Pino Toscano 15 Jul '14

15 Jul '14

Run ldconfig early in the init script, so libraries outside standard library paths but with a proper ld.so configuration file pointing at them can be found. --- appliance/init | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/appliance/init b/appliance/init index b407bf9..94e77cf 100755 --- a/appliance/init +++ b/appliance/init @@ -10,6 +10,10 @@ RUNLEVEL=S PREVLEVEL=N export RUNLEVEL PREVLEVEL +# Make sure to find all the libraries, also those in non-standard place +# but with a proper ld.so configuration pointing at them +ldconfig || : + # Try to print a stack trace for segfaults inside the appliance. for d in /lib64 /lib; do f=$d/libSegFault.so -- 1.9.3

2 3