diff options
Diffstat (limited to 'src/VBox/Additions/solaris/SharedFolders/vboxfs_vfs.c')
-rw-r--r-- | src/VBox/Additions/solaris/SharedFolders/vboxfs_vfs.c | 591 |
1 files changed, 591 insertions, 0 deletions
diff --git a/src/VBox/Additions/solaris/SharedFolders/vboxfs_vfs.c b/src/VBox/Additions/solaris/SharedFolders/vboxfs_vfs.c new file mode 100644 index 00000000..2ccf79af --- /dev/null +++ b/src/VBox/Additions/solaris/SharedFolders/vboxfs_vfs.c @@ -0,0 +1,591 @@ +/* $Id: vboxfs_vfs.c $ */ +/** @file + * VirtualBox File System for Solaris Guests, VFS implementation. + */ + +/* + * Copyright (C) 2009-2019 Oracle Corporation + * + * This file is part of VirtualBox Open Source Edition (OSE), as + * available from http://www.virtualbox.org. This file is free software; + * you can redistribute it and/or modify it under the terms of the GNU + * General Public License (GPL) as published by the Free Software + * Foundation, in version 2 as it comes in the "COPYING" file of the + * VirtualBox OSE distribution. VirtualBox OSE is distributed in the + * hope that it will be useful, but WITHOUT ANY WARRANTY of any kind. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL) only, as it comes in the "COPYING.CDDL" file of the + * VirtualBox OSE distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + */ + +#include <VBox/log.h> +#include <VBox/version.h> + +#include <sys/types.h> +#include <sys/mntent.h> +#include <sys/param.h> +#include <sys/modctl.h> +#include <sys/mount.h> +#include <sys/policy.h> +#include <sys/atomic.h> +#include <sys/sysmacros.h> +#include <sys/ddi.h> +#include <sys/sunddi.h> +#include <sys/vfs.h> +#if !defined(VBOX_VFS_SOLARIS_10U6) +# include <sys/vfs_opreg.h> +#endif +#include <sys/pathname.h> +#include <sys/cmn_err.h> +#undef u /* /usr/include/sys/user.h:249:1 is where this is defined to (curproc->p_user). very cool. */ + +#include "vboxfs_prov.h" +#include "vboxfs_vnode.h" +#include "vboxfs_vfs.h" +#include "vboxfs.h" + + +#define VBOXSOLQUOTE2(x) #x +#define VBOXSOLQUOTE(x) VBOXSOLQUOTE2(x) +/** The module name. */ +#define DEVICE_NAME "vboxfs" +/** The module description as seen in 'modinfo'. */ +#define DEVICE_DESC "VirtualBox ShrdFS" + + +/* + * Shared Folders filesystem implementation of the Solaris VFS interfaces. + * Much of this is cookie cutter code for Solaris filesystem implementation. + */ + +/* forward declarations */ +static int sffs_init(int fstype, char *name); +static int sffs_mount(vfs_t *, vnode_t *, struct mounta *, cred_t *); +static int sffs_unmount(vfs_t *vfsp, int flag, cred_t *cr); +static int sffs_root(vfs_t *vfsp, vnode_t **vpp); +static int sffs_statvfs(vfs_t *vfsp, statvfs64_t *sbp); + +static mntopt_t sffs_options[] = { + /* Option Cancels Opt Arg Flags Data */ + {"uid", NULL, NULL, MO_HASVALUE, NULL}, + {"gid", NULL, NULL, MO_HASVALUE, NULL}, + {"dmode", NULL, NULL, MO_HASVALUE, NULL}, + {"fmode", NULL, NULL, MO_HASVALUE, NULL}, + {"dmask", NULL, NULL, MO_HASVALUE, NULL}, + {"fmask", NULL, NULL, MO_HASVALUE, NULL}, + {"stat_ttl", NULL, NULL, MO_HASVALUE, NULL}, + {"fsync", NULL, NULL, 0, NULL}, + {"tag", NULL, NULL, MO_HASVALUE, NULL} +}; + +static mntopts_t sffs_options_table = { + sizeof (sffs_options) / sizeof (mntopt_t), + sffs_options +}; + +static vfsdef_t sffs_vfsdef = { + VFSDEF_VERSION, + DEVICE_NAME, + sffs_init, + VSW_HASPROTO, + &sffs_options_table +}; + +static int sffs_fstype; +static int sffs_major; /* major number for device */ + +kmutex_t sffs_minor_lock; +int sffs_minor; /* minor number for device */ + +/* + * Module linkage information + */ +static struct modlfs modlfs = { + &mod_fsops, + DEVICE_DESC " " VBOX_VERSION_STRING "r" VBOXSOLQUOTE(VBOX_SVN_REV), + &sffs_vfsdef +}; + +static struct modlinkage modlinkage = { + MODREV_1, &modlfs, NULL +}; + +static sfp_connection_t *sfprov = NULL; + +int +_init() +{ + return (mod_install(&modlinkage)); +} + +int +_info(struct modinfo *modinfop) +{ + return (mod_info(&modlinkage, modinfop)); +} + + +int +_fini() +{ + int error; + + error = mod_remove(&modlinkage); + if (error) + return (error); + + /* + * Tear down the operations vectors + */ + sffs_vnode_fini(); + (void) vfs_freevfsops_by_type(sffs_fstype); + + /* + * close connection to the provider + */ + sfprov_disconnect(sfprov); + return (0); +} + + +static int +sffs_init(int fstype, char *name) +{ +#if defined(VBOX_VFS_SOLARIS_10U6) + static const fs_operation_def_t sffs_vfsops_template[] = { + VFSNAME_MOUNT, sffs_mount, + VFSNAME_UNMOUNT, sffs_unmount, + VFSNAME_ROOT, sffs_root, + VFSNAME_STATVFS, sffs_statvfs, + NULL, NULL + }; +#else + static const fs_operation_def_t sffs_vfsops_template[] = { + VFSNAME_MOUNT, { .vfs_mount = sffs_mount }, + VFSNAME_UNMOUNT, { .vfs_unmount = sffs_unmount }, + VFSNAME_ROOT, { .vfs_root = sffs_root }, + VFSNAME_STATVFS, { .vfs_statvfs = sffs_statvfs }, + NULL, NULL + }; +#endif + int error; + + ASSERT(fstype != 0); + sffs_fstype = fstype; + LogFlowFunc(("sffs_init() name=%s\n", name)); + + /* + * This may seem a silly way to do things for now. But the code + * is structured to easily allow it to be used on other hypervisors + * which would have a different implementation of the provider. + * Hopefully that'll never happen. :) + */ + sfprov = sfprov_connect(SFPROV_VERSION); + if (sfprov == NULL) { + cmn_err(CE_WARN, "sffs_init: couldn't init sffs provider"); + return (ENODEV); + } + + error = sfprov_set_show_symlinks(); + if (error != 0) { + cmn_err(CE_WARN, "sffs_init: host unable to show symlinks, " + "rc=%d\n", error); + } + + error = vfs_setfsops(fstype, sffs_vfsops_template, NULL); + if (error != 0) { + cmn_err(CE_WARN, "sffs_init: bad vfs ops template"); + return (error); + } + + error = sffs_vnode_init(); + if (error != 0) { + (void) vfs_freevfsops_by_type(fstype); + cmn_err(CE_WARN, "sffs_init: bad vnode ops template"); + return (error); + } + + if ((sffs_major = getudev()) == (major_t)-1) { + cmn_err(CE_WARN, "sffs_init: Can't get unique device number."); + sffs_major = 0; + } + mutex_init(&sffs_minor_lock, NULL, MUTEX_DEFAULT, NULL); + return (0); +} + +/* + * wrapper for pn_get + */ +static int +sf_pn_get(char *rawpath, struct mounta *uap, char **outpath) +{ + pathname_t path; + int error; + + error = pn_get(rawpath, (uap->flags & MS_SYSSPACE) ? UIO_SYSSPACE : + UIO_USERSPACE, &path); + if (error) { + LogFlowFunc(("pn_get(%s) failed\n", rawpath)); + return (error); + } + *outpath = kmem_alloc(path.pn_pathlen + 1, KM_SLEEP); + strcpy(*outpath, path.pn_path); + pn_free(&path); + return (0); +} + +static void +sffs_print(sffs_data_t *sffs) +{ + cmn_err(CE_NOTE, "sffs_data_t at 0x%p\n", sffs); + cmn_err(CE_NOTE, " vfs_t *sf_vfsp = 0x%p\n", sffs->sf_vfsp); + cmn_err(CE_NOTE, " vnode_t *sf_rootnode = 0x%p\n", sffs->sf_rootnode); + cmn_err(CE_NOTE, " uid_t sf_uid = 0x%lu\n", (ulong_t)sffs->sf_handle->sf_uid); + cmn_err(CE_NOTE, " gid_t sf_gid = 0x%lu\n", (ulong_t)sffs->sf_handle->sf_gid); + cmn_err(CE_NOTE, " mode_t sf_dmode = 0x%lu\n", (ulong_t)sffs->sf_handle->sf_dmode); + cmn_err(CE_NOTE, " mode_t sf_fmode = 0x%lu\n", (ulong_t)sffs->sf_handle->sf_fmode); + cmn_err(CE_NOTE, " mode_t sf_dmask = 0x%lu\n", (ulong_t)sffs->sf_handle->sf_dmask); + cmn_err(CE_NOTE, " mode_t sf_fmask = 0x%lu\n", (ulong_t)sffs->sf_handle->sf_fmask); + cmn_err(CE_NOTE, " char *sf_share_name = %s\n", sffs->sf_share_name); + cmn_err(CE_NOTE, " char *sf_mntpath = %s\n", sffs->sf_mntpath); + cmn_err(CE_NOTE, " sfp_mount_t *sf_handle = 0x%p\n", sffs->sf_handle); +} + +static int +sffs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) +{ + sffs_data_t *sffs; + char *mount_point = NULL; + char *share_name = NULL; + int error; + dev_t dev; + uid_t uid = 0; + gid_t gid = 0; + mode_t dmode = ~0U; + mode_t fmode = ~0U; + mode_t dmask = 0; + mode_t fmask = 0; + int stat_ttl = DEF_STAT_TTL_MS; + int fsync = 0; + char *optval; + long val; + char *path; + sfp_mount_t *handle; + sfnode_t *sfnode; + + /* + * check we have permission to do the mount + */ + LogFlowFunc(("sffs_mount() started\n")); + error = secpolicy_fs_mount(cr, mvp, vfsp); + if (error != 0) + return (error); + + /* + * Mount point must be a directory + */ + if (mvp->v_type != VDIR) + return (ENOTDIR); + + /* + * no support for remount (what is it?) + */ + if (uap->flags & MS_REMOUNT) + return (ENOTSUP); + + /* + * Ensure that nothing else is actively in/under the mount point + */ + mutex_enter(&mvp->v_lock); + if ((uap->flags & MS_OVERLAY) == 0 && + (mvp->v_count != 1 || (mvp->v_flag & VROOT))) { + mutex_exit(&mvp->v_lock); + return (EBUSY); + } + mutex_exit(&mvp->v_lock); + + /* + * check for read only has to be done early + */ + if (uap->flags & MS_RDONLY) { + vfsp->vfs_flag |= VFS_RDONLY; + vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0); + } + + /* + * UID to use for all files + */ + if (vfs_optionisset(vfsp, "uid", &optval) && + ddi_strtol(optval, NULL, 10, &val) == 0 && + (uid_t)val == val) + uid = val; + + /* + * GID to use for all files + */ + if (vfs_optionisset(vfsp, "gid", &optval) && + ddi_strtol(optval, NULL, 10, &val) == 0 && + (gid_t)val == val) + gid = val; + + /* + * dmode to use for all directories + */ + if (vfs_optionisset(vfsp, "dmode", &optval) && + ddi_strtol(optval, NULL, 8, &val) == 0 && + (mode_t)val == val) + dmode = val; + + /* + * fmode to use for all files + */ + if (vfs_optionisset(vfsp, "fmode", &optval) && + ddi_strtol(optval, NULL, 8, &val) == 0 && + (mode_t)val == val) + fmode = val; + + /* + * dmask to use for all directories + */ + if (vfs_optionisset(vfsp, "dmask", &optval) && + ddi_strtol(optval, NULL, 8, &val) == 0 && + (mode_t)val == val) + dmask = val; + + /* + * fmask to use for all files + */ + if (vfs_optionisset(vfsp, "fmask", &optval) && + ddi_strtol(optval, NULL, 8, &val) == 0 && + (mode_t)val == val) + fmask = val; + + /* + * umask to use for all directories & files + */ + if (vfs_optionisset(vfsp, "umask", &optval) && + ddi_strtol(optval, NULL, 8, &val) == 0 && + (mode_t)val == val) + dmask = fmask = val; + + /* + * ttl to use for stat caches + */ + if (vfs_optionisset(vfsp, "stat_ttl", &optval) && + ddi_strtol(optval, NULL, 10, &val) == 0 && + (int)val == val) + { + stat_ttl = val; + } + else + vfs_setmntopt(vfsp, "stat_ttl", VBOXSOLQUOTE(DEF_STAT_TTL_MS), 0); + + /* + * whether to honor fsync + */ + if (vfs_optionisset(vfsp, "fsync", &optval)) + fsync = 1; + + /* + * Any unknown options are an error + */ + if ((uap->flags & MS_DATA) && uap->datalen > 0) { + cmn_err(CE_WARN, "sffs: unknown mount options specified"); + return (EINVAL); + } + + /* + * get the mount point pathname + */ + error = sf_pn_get(uap->dir, uap, &mount_point); + if (error) + return (error); + + /* + * find what we are mounting + */ + error = sf_pn_get(uap->spec, uap, &share_name); + if (error) { + kmem_free(mount_point, strlen(mount_point) + 1); + return (error); + } + + /* + * Invoke Hypervisor mount interface before proceeding + */ + error = sfprov_mount(sfprov, share_name, &handle); + if (error) { + kmem_free(share_name, strlen(share_name) + 1); + kmem_free(mount_point, strlen(mount_point) + 1); + return (error); + } + + /* + * find an available minor device number for this mount + */ + mutex_enter(&sffs_minor_lock); + do { + sffs_minor = (sffs_minor + 1) & L_MAXMIN32; + dev = makedevice(sffs_major, sffs_minor); + } while (vfs_devismounted(dev)); + mutex_exit(&sffs_minor_lock); + + /* + * allocate and fill in the sffs structure + */ + sffs = kmem_alloc(sizeof (*sffs), KM_SLEEP); + sffs->sf_vfsp = vfsp; + sffs->sf_handle = handle; + sffs->sf_handle->sf_uid = uid; + sffs->sf_handle->sf_gid = gid; + sffs->sf_handle->sf_dmode = dmode; + sffs->sf_handle->sf_fmode = fmode; + sffs->sf_handle->sf_dmask = dmask; + sffs->sf_handle->sf_fmask = fmask; + sffs->sf_stat_ttl = stat_ttl; + sffs->sf_fsync = fsync; + sffs->sf_share_name = share_name; + sffs->sf_mntpath = mount_point; + sffs->sf_ino = 3; /* root mount point is always '3' */ + + /* + * fill in the vfs structure + */ + vfsp->vfs_data = (caddr_t)sffs; + vfsp->vfs_fstype = sffs_fstype; + vfsp->vfs_dev = dev; + vfsp->vfs_bsize = PAGESIZE; /* HERE JOE ??? */ + vfsp->vfs_flag |= VFS_NOTRUNC; /* HERE JOE ???? */ + vfs_make_fsid(&vfsp->vfs_fsid, dev, sffs_fstype); + + /* + * create the root vnode. + * XXX JOE What should the path be here? is "/" really right? + * other options? + */ + path = kmem_alloc(2, KM_SLEEP); + strcpy(path, "."); + mutex_enter(&sffs_lock); + sfnode = sfnode_make(sffs, path, VDIR, NULL, NULL, NULL, 0); + sffs->sf_rootnode = sfnode_get_vnode(sfnode); + sffs->sf_rootnode->v_flag |= VROOT; + sffs->sf_rootnode->v_vfsp = vfsp; + mutex_exit(&sffs_lock); + + LogFlowFunc(("sffs_mount() success sffs=0x%p\n", sffs)); +#ifdef DEBUG_ramshankar + sffs_print(sffs); +#endif + return (error); +} + +static int +sffs_unmount(vfs_t *vfsp, int flag, cred_t *cr) +{ + sffs_data_t *sffs = (sffs_data_t *)vfsp->vfs_data; + int error; + + /* + * generic security check + */ + LogFlowFunc(("sffs_unmount() of sffs=0x%p\n", sffs)); + if ((error = secpolicy_fs_unmount(cr, vfsp)) != 0) + return (error); + + /* + * forced unmount is not supported by this file system + * and thus, ENOTSUP, is being returned. + */ + if (flag & MS_FORCE) { + LogFlowFunc(("sffs_unmount(MS_FORCE) returns ENOSUP\n")); + return (ENOTSUP); + } + + /* + * Mark the file system unmounted. + */ + vfsp->vfs_flag |= VFS_UNMOUNTED; + + /* + * Make sure nothing is still in use. + */ + if (sffs_purge(sffs) != 0) { + vfsp->vfs_flag &= ~VFS_UNMOUNTED; + LogFlowFunc(("sffs_unmount() returns EBUSY\n")); + return (EBUSY); + } + + /* + * Invoke Hypervisor unmount interface before proceeding + */ + error = sfprov_unmount(sffs->sf_handle); + if (error != 0) { + /* TBD anything here? */ + } + + kmem_free(sffs->sf_share_name, strlen(sffs->sf_share_name) + 1); + kmem_free(sffs->sf_mntpath, strlen(sffs->sf_mntpath) + 1); + kmem_free(sffs, sizeof(*sffs)); + LogFlowFunc(("sffs_unmount() done\n")); + return (0); +} + +/* + * return the vnode for the root of the mounted file system + */ +static int +sffs_root(vfs_t *vfsp, vnode_t **vpp) +{ + sffs_data_t *sffs = (sffs_data_t *)vfsp->vfs_data; + vnode_t *vp = sffs->sf_rootnode; + + VN_HOLD(vp); + *vpp = vp; + return (0); +} + +/* + * get some stats.. fake up the rest + */ +static int +sffs_statvfs(vfs_t *vfsp, statvfs64_t *sbp) +{ + sffs_data_t *sffs = (sffs_data_t *)vfsp->vfs_data; + sffs_fsinfo_t fsinfo; + dev32_t d32; + int error; + + bzero(sbp, sizeof(*sbp)); + error = sfprov_get_fsinfo(sffs->sf_handle, &fsinfo); + if (error != 0) + return (error); + + sbp->f_bsize = fsinfo.blksize; + sbp->f_frsize = fsinfo.blksize; + + sbp->f_bfree = fsinfo.blksavail; + sbp->f_bavail = fsinfo.blksavail; + sbp->f_files = fsinfo.blksavail / 4; /* some kind of reasonable value */ + sbp->f_ffree = fsinfo.blksavail / 4; + sbp->f_favail = fsinfo.blksavail / 4; + + sbp->f_blocks = fsinfo.blksused + sbp->f_bavail; + + (void) cmpldev(&d32, vfsp->vfs_dev); + sbp->f_fsid = d32; + strcpy(&sbp->f_basetype[0], "sffs"); + sbp->f_flag |= ST_NOSUID; + + if (fsinfo.readonly) + sbp->f_flag |= ST_RDONLY; + + sbp->f_namemax = fsinfo.maxnamesize; + return (0); +} + |