collabora-online/tools/mount.cpp

342 lines
10 KiB
C++
Raw Permalink Normal View History

/* -*- Mode: C; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4; fill-column: 100 -*- */
/*
* Copyright the Collabora Online contributors.
*
* SPDX-License-Identifier: MPL-2.0
*
* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/.
*/
/*
* This is a very tiny helper to allow overlay mounting.
*/
#include <config.h>
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
#include <stdio.h>
#include <errno.h>
#include <sys/mount.h>
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
#include <sys/stat.h>
#include <sysexits.h>
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
#include <unistd.h>
#include <security.h>
#ifdef __FreeBSD__
#include <stdlib.h>
#include <sys/uio.h>
#define MOUNT mount_wrapper
#define MS_MGC_VAL 0
#define MS_SILENT 0
#define MS_NODEV 0
#define MS_UNBINDABLE 0
#define MS_BIND 1
#define MS_REC 2
#define MS_REMOUNT 4
#define MS_NOATIME 8
#define MS_NOSUID 16
#define MS_RDONLY 32
void
build_iovec(struct iovec **iov, int *iovlen, const char *name, const void *val,
size_t len)
{
int i;
if (*iovlen < 0)
return;
i = *iovlen;
*iov = reinterpret_cast<struct iovec*>(realloc(*iov, sizeof **iov * (i + 2)));
if (*iov == NULL) {
*iovlen = -1;
return;
}
(*iov)[i].iov_base = strdup(name);
(*iov)[i].iov_len = strlen(name) + 1;
i++;
(*iov)[i].iov_base = const_cast<void*>(val);
if (len == (size_t)-1) {
if (val != NULL)
len = strlen(reinterpret_cast<const char*>(val)) + 1;
else
len = 0;
}
(*iov)[i].iov_len = (int)len;
*iovlen = ++i;
}
int mount_wrapper(const char *source, const char *target,
const char *filesystemtype, unsigned long mountflags,
const void *data)
{
struct iovec *iov = NULL;
int iovlen = 0;
int freebsd_flags = 0;
(void)data;
if(mountflags & MS_BIND)
filesystemtype = "nullfs";
if(mountflags & MS_REMOUNT)
freebsd_flags |= MNT_UPDATE;
if(mountflags & MS_NOATIME)
freebsd_flags |= MNT_NOATIME;
if(mountflags & MS_NOSUID)
freebsd_flags |= MNT_NOSUID;
if(mountflags & MS_RDONLY)
freebsd_flags |= MNT_RDONLY;
// TODO: handle MS_REC?
build_iovec(&iov, &iovlen, "fstype", reinterpret_cast<const void*>(filesystemtype), (size_t)-1);
build_iovec(&iov, &iovlen, "fspath", reinterpret_cast<const void*>(target), (size_t)-1);
build_iovec(&iov, &iovlen, "from", reinterpret_cast<const void*>(source), (size_t)-1);
return nmount(iov, iovlen, freebsd_flags);
}
#define MNT_DETACH 1
int umount2(const char *target, int flags)
{
struct statfs* mntbufs;
int numInfos = getmntinfo(&mntbufs, MNT_WAIT);
bool targetMounted = false;
for (int i = 0; i < numInfos; i++)
{
if (!strcmp(target, mntbufs[i].f_mntonname))
{
targetMounted = true;
break;
}
}
if (!targetMounted)
{
errno = EINVAL;
return -1;
}
if(flags == MNT_DETACH)
flags = 0;
return unmount(target, flags);
}
#else
#define MOUNT mount
#endif
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
void usage(const char* program)
{
fprintf(stderr, "Usage: %s <-b|-r> <source path> <target path>\n", program);
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
fprintf(stderr, " %s -u <target>.\n", program);
#ifdef __FreeBSD__
fprintf(stderr, " %s -d <target>.\n", program);
#endif
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
fprintf(stderr, " -b bind and mount the source to target.\n");
fprintf(stderr, " -r bind and mount the source to target as readonly.\n");
#ifdef __FreeBSD__
fprintf(stderr, " -d mount minimal devfs layout (random and urandom) to target.\n");
#endif
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
fprintf(stderr, " -u to unmount the target.\n");
}
int main(int argc, char** argv)
{
/*WARNING: PRIVILEGED CODE CHECKING START */
/*WARNING*/ if (!hasCorrectUID(/* appName = */ "coolmount"))
/*WARNING*/ {
/*WARNING*/ fprintf(stderr, "Aborting.\n");
/*WARNING*/ return EX_SOFTWARE;
/*WARNING*/ }
/*WARNING: PRIVILEGED CODE CHECKING END */
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
const char* program = argv[0];
if (argc < 3)
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
{
usage(program);
return EX_USAGE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
const char* option = argv[1];
if (argc == 3 && strcmp(option, "-u") == 0) // Unmount
{
const char* target = argv[2];
struct stat sb;
const bool target_exists = (stat(target, &sb) == 0 && S_ISDIR(sb.st_mode));
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
// Do nothing if target doesn't exist.
if (target_exists)
{
// Unmount the target, first by detaching. This should succeed.
int retval = umount2(target, MNT_DETACH);
if (retval != 0)
{
if (errno != EINVAL)
{
// Don't complain where MNT_DETACH is unsupported. Fall back to forcing instead.
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
fprintf(stderr, "%s: unmount failed to detach [%s]: %s.\n", program, target,
strerror(errno));
}
// Now try to force the unmounting, which isn't supported on all filesystems.
retval = umount2(target, MNT_FORCE);
if (retval != 0)
{
// From man umount(2), MNT_FORCE is not commonly supported:
// As at Linux 4.12, MNT_FORCE is supported only on the following filesystems: 9p (since
// Linux 2.6.16), ceph (since Linux 2.6.34), cifs (since Linux 2.6.12),
// fuse (since Linux 2.6.16), lustre (since Linux 3.11), and NFS (since Linux 2.1.116).
if (errno != EINVAL)
{
// Complain to capture the reason of failure.
fprintf(stderr, "%s: forced unmount of [%s] failed: %s.\n", program, target,
strerror(errno));
}
return EX_SOFTWARE;
}
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
}
}
#ifdef __FreeBSD__
else if (argc == 3 && strcmp(option, "-d") == 0) // Mount devfs
{
const char* target = argv[2];
struct stat sb;
const bool target_exists = (stat(target, &sb) == 0 && S_ISDIR(sb.st_mode));
if (!target_exists)
{
fprintf(stderr, "%s: cannot mount on invalid target directory [%s].\n", program,
target);
return EX_USAGE;
}
struct iovec *iov = NULL;
int iovlen = 0;
build_iovec(&iov, &iovlen, "fstype", "devfs", (size_t)-1);
build_iovec(&iov, &iovlen, "fspath", reinterpret_cast<const void*>(target), (size_t)-1);
build_iovec(&iov, &iovlen, "from", "devfs", (size_t)-1);
// See /etc/defaults/devfs.rules
// [devfsrules_jail=4]
build_iovec(&iov, &iovlen, "ruleset", "4", (size_t)-1);
int retval = nmount(iov, iovlen, 0);
if (retval)
{
fprintf(stderr, "%s: mount failed create to devfs layout in [%s]: %s.\n", program, target,
strerror(errno));
return EX_SOFTWARE;
}
}
#endif
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
else if (argc == 4) // Mount
{
const char* source = argv[2];
struct stat sb;
if (stat(source, &sb))
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
{
fprintf(stderr, "%s: cannot mount from invalid source [%s]. stat failed with %s\n",
program, source, strerror(errno));
return EX_USAGE;
}
const bool isDir = S_ISDIR(sb.st_mode);
const bool isCharDev = S_ISCHR(sb.st_mode); // We don't support regular files.
if (isCharDev)
{
// Even for character devices, we only support the random devices.
if (strstr("/dev/random", source) && strstr("/dev/urandom", source))
{
fprintf(stderr, "%s: cannot mount untrusted character-device [%s]", program,
source);
return EX_USAGE;
}
}
if (!isDir && !isCharDev)
{
fprintf(stderr,
"%s: cannot mount from invalid source [%s], it is neither a file nor a "
"directory.\n",
program, source);
return EX_USAGE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
const char* target = argv[3];
if (stat(target, &sb))
{
fprintf(stderr, "%s: cannot mount on invalid target [%s]. stat failed with %s\n",
program, target, strerror(errno));
return EX_USAGE;
}
const bool target_exists =
((isDir && S_ISDIR(sb.st_mode)) || (isCharDev && S_ISREG(sb.st_mode)));
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
if (!target_exists)
{
fprintf(stderr,
"%s: cannot mount on invalid target [%s], it is not a %s as the source\n",
program, target, isDir ? "directory" : "file");
return EX_USAGE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
// Mount the source path as the target path.
// First bind to mount an existing directory node into the chroot.
// MS_BIND ignores other flags.
if (strcmp(option, "-b") == 0) // Shared or Bind Mount.
{
const int retval
= MOUNT(source, target, nullptr, (MS_MGC_VAL | MS_BIND | MS_REC), nullptr);
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
if (retval)
{
fprintf(stderr, "%s: mount failed to bind [%s] to [%s]: %s.\n", program, source,
target, strerror(errno));
return EX_SOFTWARE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
}
else if (strcmp(option, "-r") == 0) // Readonly Mount.
{
// Now we need to set read-only and other flags with a remount.
int retval = MOUNT(source, target, nullptr,
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
(MS_BIND | MS_REC | MS_REMOUNT | MS_NOATIME | MS_NODEV | MS_NOSUID
| MS_RDONLY | MS_SILENT),
nullptr);
if (retval)
{
fprintf(stderr, "%s: mount failed remount [%s] readonly: %s.\n", program, target,
strerror(errno));
return EX_SOFTWARE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
retval = MOUNT(source, target, nullptr, (MS_UNBINDABLE | MS_REC), nullptr);
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
if (retval)
{
fprintf(stderr, "%s: mount failed make [%s] private: %s.\n", program, target,
strerror(errno));
return EX_SOFTWARE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
}
}
else
{
usage(program);
return EX_USAGE;
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
}
wsd: faster jail setup via bind-mount loolmount now works and supports mounting and unmounting, plus numerous improvements, refactoring, logging, etc.. When enabled, binding improves the jail setup time by anywhere from 2x to orders of magnitude (in docker, f.e.). A new config entry mount_jail_tree controls whether mounting is used or the old method of linking/copying of jail contents. It is set to true by default and falls back to linking/copying. A test mount is done when the setting is enabled, and if mounting fails, it's disabled to avoid noise. Temporarily disabled for unit-tests until we can cleanup lingering mounts after Jenkins aborts our build job. In a future patch we will have mount/jail cleanup as part of make. The network/system files in /etc that need frequent refreshing are now updated in systemplate to make their most recent version available in the jails. These files can change during the course of loolwsd lifetime, and are unlikely to be updated in systemplate after installation at all. We link to them in the systemplate/etc directory, and if that fails, we copy them before forking each kit instance to have the latest. This reworks the approach used to bind-mount the jails and the templates such that the total is now down to only three mounts: systemplate, lo, tmp. As now systemplate and lotemplate are shared, they must be mounted as readonly, this means that user/ must now be moved into tmp/user/ which is writable. The mount-points must be recursive, because we mount lo/ within the mount-point of systemplate (which is the root of the jail). But because we (re)bind recursively, and because both systemplate and lotemplate are mounted for each jails, we need to make them unbindable, so they wouldn't multiply the mount-points for each jails (an explosive growth!) Contrarywise, we don't want the mount-points to be shared, because we don't expect to add/remove mounts after a jail is created. The random temp directory is now created and set correctly, plus many logging and other improvements. Change-Id: Iae3fda5e876cf47d2cae6669a87b5b826a8748df Reviewed-on: https://gerrit.libreoffice.org/c/online/+/92829 Tested-by: Jenkins Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2020-04-09 15:02:58 +02:00
fflush(stderr);
return EX_OK;
}
/* vim:set shiftwidth=4 softtabstop=4 expandtab: */