2012-02-13 16:28:04 +01:00
|
|
|
/* lookup() is the main routine that controls the path name lookup. It
|
2006-10-25 15:40:36 +02:00
|
|
|
* handles mountpoints and symbolic links. The actual lookup requests
|
|
|
|
* are sent through the req_lookup wrapper function.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "fs.h"
|
|
|
|
#include <string.h>
|
|
|
|
#include <minix/callnr.h>
|
|
|
|
#include <minix/com.h>
|
|
|
|
#include <minix/keymap.h>
|
|
|
|
#include <minix/const.h>
|
|
|
|
#include <minix/endpoint.h>
|
|
|
|
#include <unistd.h>
|
2010-03-29 13:39:54 +02:00
|
|
|
#include <assert.h>
|
2006-10-25 15:40:36 +02:00
|
|
|
#include <minix/vfsif.h>
|
2010-08-30 15:44:07 +02:00
|
|
|
#include <sys/stat.h>
|
|
|
|
#include <sys/un.h>
|
|
|
|
#include <dirent.h>
|
2012-02-13 16:28:04 +01:00
|
|
|
#include "threads.h"
|
2006-10-25 15:40:36 +02:00
|
|
|
#include "vmnt.h"
|
|
|
|
#include "vnode.h"
|
2012-02-13 16:28:04 +01:00
|
|
|
#include "path.h"
|
|
|
|
#include "fproc.h"
|
2006-10-25 15:40:36 +02:00
|
|
|
#include "param.h"
|
|
|
|
|
2007-08-08 13:40:47 +02:00
|
|
|
/* Set to following define to 1 if you really want to use the POSIX definition
|
|
|
|
* (IEEE Std 1003.1, 2004) of pathname resolution. POSIX requires pathnames
|
|
|
|
* with a traling slash (and that do not entirely consist of slash characters)
|
|
|
|
* to be treated as if a single dot is appended. This means that for example
|
|
|
|
* mkdir("dir/", ...) and rmdir("dir/") will fail because the call tries to
|
|
|
|
* create or remove the directory '.'. Historically, Unix systems just ignore
|
|
|
|
* trailing slashes.
|
|
|
|
*/
|
|
|
|
#define DO_POSIX_PATHNAME_RES 0
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
FORWARD _PROTOTYPE( int lookup, (struct vnode *dirp, struct lookup *resolve,
|
2010-08-30 15:44:07 +02:00
|
|
|
node_details_t *node, struct fproc *rfp));
|
2012-02-13 16:28:04 +01:00
|
|
|
FORWARD _PROTOTYPE( int check_perms, (endpoint_t ep, cp_grant_id_t io_gr,
|
|
|
|
size_t pathlen) );
|
2007-08-07 14:52:47 +02:00
|
|
|
|
2006-10-25 15:40:36 +02:00
|
|
|
/*===========================================================================*
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
* advance *
|
2006-10-25 15:40:36 +02:00
|
|
|
*===========================================================================*/
|
2012-02-13 16:28:04 +01:00
|
|
|
PUBLIC struct vnode *advance(dirp, resolve, rfp)
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
struct vnode *dirp;
|
2012-02-13 16:28:04 +01:00
|
|
|
struct lookup *resolve;
|
2010-08-30 15:44:07 +02:00
|
|
|
struct fproc *rfp;
|
2006-10-25 15:40:36 +02:00
|
|
|
{
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Resolve a path name starting at dirp to a vnode. */
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
int r;
|
2012-02-13 16:28:04 +01:00
|
|
|
int do_downgrade = 1;
|
2007-08-07 14:52:47 +02:00
|
|
|
struct vnode *new_vp, *vp;
|
2006-10-25 15:40:36 +02:00
|
|
|
struct vmnt *vmp;
|
2010-11-23 20:34:56 +01:00
|
|
|
struct node_details res = {0,0,0,0,0,0,0};
|
2012-02-13 16:28:04 +01:00
|
|
|
tll_access_t initial_locktype;
|
2006-10-25 15:40:36 +02:00
|
|
|
|
2010-03-29 13:39:54 +02:00
|
|
|
assert(dirp);
|
2012-02-13 16:28:04 +01:00
|
|
|
assert(resolve->l_vnode_lock != TLL_NONE);
|
|
|
|
assert(resolve->l_vmnt_lock != TLL_NONE);
|
|
|
|
|
|
|
|
if (resolve->l_vnode_lock == VNODE_READ)
|
|
|
|
initial_locktype = VNODE_OPCL;
|
|
|
|
else
|
|
|
|
initial_locktype = resolve->l_vnode_lock;
|
|
|
|
|
|
|
|
/* Get a free vnode and lock it */
|
|
|
|
if ((new_vp = get_free_vnode()) == NULL) return(NULL);
|
|
|
|
lock_vnode(new_vp, initial_locktype);
|
2010-03-29 13:39:54 +02:00
|
|
|
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
/* Lookup vnode belonging to the file. */
|
2012-02-13 16:28:04 +01:00
|
|
|
if ((r = lookup(dirp, resolve, &res, rfp)) != OK) {
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
err_code = r;
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(new_vp);
|
2010-05-10 15:26:00 +02:00
|
|
|
return(NULL);
|
2007-08-07 14:52:47 +02:00
|
|
|
}
|
2012-02-13 16:28:04 +01:00
|
|
|
|
|
|
|
/* Check whether we already have a vnode for that file */
|
2010-05-10 15:26:00 +02:00
|
|
|
if ((vp = find_vnode(res.fs_e, res.inode_nr)) != NULL) {
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(new_vp); /* Don't need this anymore */
|
|
|
|
do_downgrade = (lock_vnode(vp, initial_locktype) != EBUSY);
|
|
|
|
|
|
|
|
/* Unfortunately, by the time we get the lock, another thread might've
|
|
|
|
* rid of the vnode (e.g., find_vnode found the vnode while a
|
|
|
|
* req_putnode was being processed). */
|
|
|
|
if (vp->v_ref_count == 0) { /* vnode vanished! */
|
|
|
|
/* As the lookup before increased the usage counters in the FS,
|
|
|
|
* we can simply set the usage counters to 1 and proceed as
|
|
|
|
* normal, because the putnode resulted in a use count of 1 in
|
|
|
|
* the FS. Other data is still valid, because the vnode was
|
|
|
|
* marked as pending lock, so get_free_vnode hasn't
|
|
|
|
* reinitialized the vnode yet. */
|
|
|
|
vp->v_fs_count = 1;
|
|
|
|
if (vp->v_mapfs_e != NONE) vp->v_mapfs_count = 1;
|
|
|
|
} else {
|
|
|
|
vp->v_fs_count++; /* We got a reference from the FS */
|
|
|
|
}
|
|
|
|
|
|
|
|
} else {
|
|
|
|
/* Vnode not found, fill in the free vnode's fields */
|
|
|
|
|
|
|
|
new_vp->v_fs_e = res.fs_e;
|
|
|
|
new_vp->v_inode_nr = res.inode_nr;
|
|
|
|
new_vp->v_mode = res.fmode;
|
|
|
|
new_vp->v_size = res.fsize;
|
|
|
|
new_vp->v_uid = res.uid;
|
|
|
|
new_vp->v_gid = res.gid;
|
|
|
|
new_vp->v_sdev = res.dev;
|
|
|
|
|
|
|
|
if( (vmp = find_vmnt(new_vp->v_fs_e)) == NULL)
|
|
|
|
panic("advance: vmnt not found");
|
|
|
|
|
|
|
|
new_vp->v_vmnt = vmp;
|
|
|
|
new_vp->v_dev = vmp->m_dev;
|
|
|
|
new_vp->v_fs_count = 1;
|
|
|
|
|
|
|
|
vp = new_vp;
|
2007-08-07 14:52:47 +02:00
|
|
|
}
|
2006-10-25 15:40:36 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
dup_vnode(vp);
|
|
|
|
if (do_downgrade) {
|
|
|
|
/* Only downgrade a lock if we managed to lock it in the first place */
|
|
|
|
*(resolve->l_vnode) = vp;
|
|
|
|
|
|
|
|
if (initial_locktype != resolve->l_vnode_lock)
|
|
|
|
tll_downgrade(&vp->v_lock);
|
|
|
|
|
|
|
|
#if LOCK_DEBUG
|
|
|
|
if (resolve->l_vnode_lock == VNODE_READ)
|
|
|
|
fp->fp_vp_rdlocks++;
|
|
|
|
#endif
|
|
|
|
}
|
2006-10-25 15:40:36 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
return(vp);
|
|
|
|
}
|
2006-10-25 15:40:36 +02:00
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
/*===========================================================================*
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
* eat_path *
|
2007-08-07 14:52:47 +02:00
|
|
|
*===========================================================================*/
|
2012-02-13 16:28:04 +01:00
|
|
|
PUBLIC struct vnode *eat_path(resolve, rfp)
|
|
|
|
struct lookup *resolve;
|
2010-08-30 15:44:07 +02:00
|
|
|
struct fproc *rfp;
|
2007-08-07 14:52:47 +02:00
|
|
|
{
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Resolve path to a vnode. advance does the actual work. */
|
|
|
|
struct vnode *start_dir;
|
2006-10-25 15:40:36 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
start_dir = (resolve->l_path[0] == '/' ? rfp->fp_rd : rfp->fp_wd);
|
|
|
|
return advance(start_dir, resolve, rfp);
|
2007-01-05 17:36:55 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*===========================================================================*
|
2010-08-30 15:44:07 +02:00
|
|
|
* last_dir *
|
2007-01-05 17:36:55 +01:00
|
|
|
*===========================================================================*/
|
2012-02-13 16:28:04 +01:00
|
|
|
PUBLIC struct vnode *last_dir(resolve, rfp)
|
|
|
|
struct lookup *resolve;
|
2010-08-30 15:44:07 +02:00
|
|
|
struct fproc *rfp;
|
2007-01-05 17:36:55 +01:00
|
|
|
{
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Parse a path, as far as the last directory, fetch the vnode
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
* for the last directory into the vnode table, and return a pointer to the
|
|
|
|
* vnode. In addition, return the final component of the path in 'string'. If
|
2010-05-10 15:26:00 +02:00
|
|
|
* the last directory can't be opened, return NULL and the reason for
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
* failure in 'err_code'. We can't parse component by component as that would
|
|
|
|
* be too expensive. Alternatively, we cut off the last component of the path,
|
|
|
|
* and parse the path up to the penultimate component.
|
2012-02-13 16:28:04 +01:00
|
|
|
*/
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
|
|
|
|
size_t len;
|
|
|
|
char *cp;
|
2011-09-12 11:00:24 +02:00
|
|
|
char dir_entry[NAME_MAX+1];
|
2012-02-13 16:28:04 +01:00
|
|
|
struct vnode *start_dir, *res_vp, *sym_vp, *loop_start;
|
|
|
|
struct vmnt *sym_vmp = NULL;
|
|
|
|
int r, symloop = 0, ret_on_symlink = 0;
|
|
|
|
struct lookup symlink;
|
2007-08-07 14:52:47 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
*resolve->l_vnode = NULL;
|
|
|
|
*resolve->l_vmp = NULL;
|
|
|
|
loop_start = NULL;
|
|
|
|
sym_vp = NULL;
|
2007-08-07 14:52:47 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
ret_on_symlink = !!(resolve->l_flags & PATH_RET_SYMLINK);
|
|
|
|
|
|
|
|
do {
|
|
|
|
/* Is the path absolute or relative? Initialize 'start_dir'
|
|
|
|
* accordingly. Use loop_start in case we're looping.
|
|
|
|
*/
|
|
|
|
if (loop_start != NULL)
|
|
|
|
start_dir = loop_start;
|
|
|
|
else
|
|
|
|
start_dir = (resolve->l_path[0] == '/' ? rfp->fp_rd:rfp->fp_wd);
|
|
|
|
|
|
|
|
len = strlen(resolve->l_path);
|
|
|
|
|
|
|
|
/* If path is empty, return ENOENT. */
|
|
|
|
if (len == 0) {
|
|
|
|
err_code = ENOENT;
|
|
|
|
res_vp = NULL;
|
|
|
|
break;
|
|
|
|
}
|
2007-08-07 14:52:47 +02:00
|
|
|
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
#if !DO_POSIX_PATHNAME_RES
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Remove trailing slashes */
|
|
|
|
while (len > 1 && resolve->l_path[len-1] == '/') {
|
|
|
|
len--;
|
|
|
|
resolve->l_path[len]= '\0';
|
|
|
|
}
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
#endif
|
2007-01-05 17:36:55 +01:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
cp = strrchr(resolve->l_path, '/');
|
|
|
|
if (cp == NULL) {
|
2012-02-21 11:16:42 +01:00
|
|
|
/* Just an entry in the current working directory. Prepend
|
|
|
|
* "./" in front of the path and resolve it.
|
|
|
|
*/
|
|
|
|
strncpy(dir_entry, resolve->l_path, NAME_MAX);
|
|
|
|
dir_entry[NAME_MAX] = '\0';
|
|
|
|
resolve->l_path[0] = '.';
|
|
|
|
resolve->l_path[1] = '\0';
|
2012-02-13 16:28:04 +01:00
|
|
|
} else if (cp[1] == '\0') {
|
|
|
|
/* Path ends in a slash. The directory entry is '.' */
|
|
|
|
strcpy(dir_entry, ".");
|
|
|
|
} else {
|
|
|
|
/* A path name for the directory and a directory entry */
|
|
|
|
strncpy(dir_entry, cp+1, NAME_MAX);
|
|
|
|
cp[1] = '\0';
|
|
|
|
dir_entry[NAME_MAX] = '\0';
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Remove trailing slashes */
|
|
|
|
while (cp > resolve->l_path && cp[0] == '/') {
|
|
|
|
cp[0]= '\0';
|
|
|
|
cp--;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Resolve up to and including the last directory of the path. Turn off
|
|
|
|
* PATH_RET_SYMLINK, because we do want to follow the symlink in this
|
|
|
|
* case. That is, the flag is meant for the actual filename of the path,
|
|
|
|
* not the last directory.
|
|
|
|
*/
|
|
|
|
resolve->l_flags &= ~PATH_RET_SYMLINK;
|
|
|
|
if ((res_vp = advance(start_dir, resolve, rfp)) == NULL) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If the directory entry is not a symlink we're done now. If it is a
|
|
|
|
* symlink, then we're not at the last directory, yet. */
|
|
|
|
|
|
|
|
/* Copy the directory entry back to user_fullpath */
|
|
|
|
strncpy(resolve->l_path, dir_entry, NAME_MAX + 1);
|
|
|
|
|
|
|
|
/* Look up the directory entry, but do not follow the symlink when it
|
|
|
|
* is one.
|
|
|
|
*/
|
|
|
|
lookup_init(&symlink, resolve->l_path,
|
|
|
|
resolve->l_flags|PATH_RET_SYMLINK, &sym_vmp, &sym_vp);
|
|
|
|
symlink.l_vnode_lock = VNODE_READ;
|
|
|
|
symlink.l_vmnt_lock = VMNT_READ;
|
|
|
|
sym_vp = advance(res_vp, &symlink, rfp);
|
|
|
|
|
|
|
|
if (sym_vp != NULL && S_ISLNK(sym_vp->v_mode)) {
|
|
|
|
/* Last component is a symlink, but if we've been asked to not
|
|
|
|
* resolve it, return now.
|
|
|
|
*/
|
|
|
|
if (ret_on_symlink) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
r = req_rdlink(sym_vp->v_fs_e, sym_vp->v_inode_nr, NONE,
|
|
|
|
resolve->l_path, PATH_MAX - 1, 1);
|
|
|
|
|
|
|
|
if (r < 0) {
|
|
|
|
/* Failed to read link */
|
|
|
|
err_code = r;
|
|
|
|
unlock_vnode(res_vp);
|
|
|
|
unlock_vmnt(*resolve->l_vmp);
|
|
|
|
put_vnode(res_vp);
|
|
|
|
*resolve->l_vmp = NULL;
|
|
|
|
*resolve->l_vnode = NULL;
|
|
|
|
res_vp = NULL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
resolve->l_path[r] = '\0';
|
|
|
|
|
|
|
|
if (strrchr(resolve->l_path, '/') != NULL) {
|
|
|
|
unlock_vnode(sym_vp);
|
|
|
|
unlock_vmnt(*resolve->l_vmp);
|
|
|
|
if (sym_vmp != NULL)
|
|
|
|
unlock_vmnt(sym_vmp);
|
|
|
|
*resolve->l_vmp = NULL;
|
|
|
|
put_vnode(sym_vp);
|
|
|
|
sym_vp = NULL;
|
|
|
|
|
|
|
|
symloop++;
|
|
|
|
|
|
|
|
/* Relative symlinks are relative to res_vp, not cwd */
|
|
|
|
if (resolve->l_path[0] != '/') {
|
|
|
|
loop_start = res_vp;
|
|
|
|
} else {
|
|
|
|
/* Absolute symlink, forget about res_vp */
|
|
|
|
unlock_vnode(res_vp);
|
|
|
|
put_vnode(res_vp);
|
|
|
|
}
|
|
|
|
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
} while (symloop < SYMLOOP_MAX);
|
|
|
|
|
|
|
|
if (symloop >= SYMLOOP_MAX) {
|
|
|
|
err_code = ELOOP;
|
|
|
|
res_vp = NULL;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
}
|
2007-08-07 14:52:47 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (sym_vp != NULL) {
|
|
|
|
unlock_vnode(sym_vp);
|
|
|
|
if (sym_vmp != NULL) {
|
|
|
|
unlock_vmnt(sym_vmp);
|
|
|
|
}
|
|
|
|
put_vnode(sym_vp);
|
2009-04-29 18:59:18 +02:00
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (loop_start != NULL) {
|
|
|
|
unlock_vnode(loop_start);
|
|
|
|
put_vnode(loop_start);
|
|
|
|
}
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
|
|
|
|
/* Copy the directory entry back to user_fullpath */
|
2012-02-13 16:28:04 +01:00
|
|
|
strncpy(resolve->l_path, dir_entry, NAME_MAX + 1);
|
2012-02-21 11:16:42 +01:00
|
|
|
|
|
|
|
/* Turn PATH_RET_SYMLINK flag back on if it was on */
|
|
|
|
if (ret_on_symlink) resolve->l_flags |= PATH_RET_SYMLINK;
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
return(res_vp);
|
2007-08-07 14:52:47 +02:00
|
|
|
}
|
2007-01-05 17:36:55 +01:00
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
/*===========================================================================*
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
* lookup *
|
2007-08-07 14:52:47 +02:00
|
|
|
*===========================================================================*/
|
2012-02-13 16:28:04 +01:00
|
|
|
PRIVATE int lookup(start_node, resolve, result_node, rfp)
|
2007-08-07 14:52:47 +02:00
|
|
|
struct vnode *start_node;
|
2012-02-13 16:28:04 +01:00
|
|
|
struct lookup *resolve;
|
|
|
|
node_details_t *result_node;
|
2010-08-30 15:44:07 +02:00
|
|
|
struct fproc *rfp;
|
2007-08-07 14:52:47 +02:00
|
|
|
{
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Resolve a path name relative to start_node. */
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
int r, symloop;
|
|
|
|
endpoint_t fs_e;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
size_t path_off, path_left_len;
|
2007-08-07 14:52:47 +02:00
|
|
|
ino_t dir_ino, root_ino;
|
|
|
|
uid_t uid;
|
|
|
|
gid_t gid;
|
|
|
|
struct vnode *dir_vp;
|
2012-02-13 16:28:04 +01:00
|
|
|
struct vmnt *vmp, *vmpres;
|
2007-08-07 14:52:47 +02:00
|
|
|
struct lookup_res res;
|
2009-09-21 16:49:26 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
assert(resolve->l_vmp);
|
|
|
|
assert(resolve->l_vnode);
|
|
|
|
|
|
|
|
*(resolve->l_vmp) = vmpres = NULL; /* No vmnt found nor locked yet */
|
|
|
|
|
2007-01-05 17:36:55 +01:00
|
|
|
/* Empty (start) path? */
|
2012-02-13 16:28:04 +01:00
|
|
|
if (resolve->l_path[0] == '\0') {
|
|
|
|
result_node->inode_nr = 0;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(ENOENT);
|
2007-01-05 17:36:55 +01:00
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (!rfp->fp_rd || !rfp->fp_wd) {
|
|
|
|
printf("VFS: lookup %d: no rd/wd\n", rfp->fp_endpoint);
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(ENOENT);
|
2009-04-29 18:59:18 +02:00
|
|
|
}
|
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
fs_e = start_node->v_fs_e;
|
|
|
|
dir_ino = start_node->v_inode_nr;
|
2012-02-13 16:28:04 +01:00
|
|
|
vmpres = find_vmnt(fs_e);
|
|
|
|
|
|
|
|
if (vmpres == NULL) return(EIO); /* mountpoint vanished? */
|
|
|
|
|
2007-01-05 17:36:55 +01:00
|
|
|
/* Is the process' root directory on the same partition?,
|
|
|
|
* if so, set the chroot directory too. */
|
2010-08-30 15:44:07 +02:00
|
|
|
if (rfp->fp_rd->v_dev == rfp->fp_wd->v_dev)
|
2012-02-13 16:28:04 +01:00
|
|
|
root_ino = rfp->fp_rd->v_inode_nr;
|
2007-01-05 17:36:55 +01:00
|
|
|
else
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
root_ino = 0;
|
2007-01-05 17:36:55 +01:00
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
/* Set user and group ids according to the system call */
|
2012-02-13 16:28:04 +01:00
|
|
|
uid = (call_nr == ACCESS ? rfp->fp_realuid : rfp->fp_effuid);
|
|
|
|
gid = (call_nr == ACCESS ? rfp->fp_realgid : rfp->fp_effgid);
|
2007-08-07 14:52:47 +02:00
|
|
|
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
symloop = 0; /* Number of symlinks seen so far */
|
2007-01-05 17:36:55 +01:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Lock vmnt */
|
|
|
|
if ((r = lock_vmnt(vmpres, resolve->l_vmnt_lock)) != OK) {
|
|
|
|
if (r == EBUSY) /* vmnt already locked */
|
|
|
|
vmpres = NULL;
|
|
|
|
else
|
|
|
|
return(r);
|
|
|
|
}
|
|
|
|
*(resolve->l_vmp) = vmpres;
|
|
|
|
|
2007-01-05 17:36:55 +01:00
|
|
|
/* Issue the request */
|
2012-02-13 16:28:04 +01:00
|
|
|
r = req_lookup(fs_e, dir_ino, root_ino, uid, gid, resolve, &res, rfp);
|
2007-08-07 14:52:47 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (r != OK && r != EENTERMOUNT && r != ELEAVEMOUNT && r != ESYMLINK) {
|
|
|
|
if (vmpres) unlock_vmnt(vmpres);
|
|
|
|
*(resolve->l_vmp) = NULL;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(r); /* i.e., an error occured */
|
2012-02-13 16:28:04 +01:00
|
|
|
}
|
2007-01-05 17:36:55 +01:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* While the response is related to mount control set the
|
2007-01-05 17:36:55 +01:00
|
|
|
* new requests respectively */
|
2012-02-13 16:28:04 +01:00
|
|
|
while (r == EENTERMOUNT || r == ELEAVEMOUNT || r == ESYMLINK) {
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
/* Update user_fullpath to reflect what's left to be parsed. */
|
|
|
|
path_off = res.char_processed;
|
2012-02-13 16:28:04 +01:00
|
|
|
path_left_len = strlen(&resolve->l_path[path_off]);
|
|
|
|
memmove(resolve->l_path, &resolve->l_path[path_off], path_left_len);
|
|
|
|
resolve->l_path[path_left_len] = '\0'; /* terminate string */
|
2007-08-07 14:52:47 +02:00
|
|
|
|
|
|
|
/* Update the current value of the symloop counter */
|
|
|
|
symloop += res.symloop;
|
2012-02-13 16:28:04 +01:00
|
|
|
if (symloop > SYMLOOP_MAX) {
|
|
|
|
if (vmpres) unlock_vmnt(vmpres);
|
|
|
|
*(resolve->l_vmp) = NULL;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(ELOOP);
|
2012-02-13 16:28:04 +01:00
|
|
|
}
|
2007-08-07 14:52:47 +02:00
|
|
|
|
|
|
|
/* Symlink encountered with absolute path */
|
|
|
|
if (r == ESYMLINK) {
|
2010-08-30 15:44:07 +02:00
|
|
|
dir_vp = rfp->fp_rd;
|
2012-02-13 16:28:04 +01:00
|
|
|
vmp = NULL;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
} else if (r == EENTERMOUNT) {
|
2007-08-07 14:52:47 +02:00
|
|
|
/* Entering a new partition */
|
2012-02-13 16:28:04 +01:00
|
|
|
dir_vp = NULL;
|
2007-08-07 14:52:47 +02:00
|
|
|
/* Start node is now the mounted partition's root node */
|
|
|
|
for (vmp = &vmnt[0]; vmp != &vmnt[NR_MNTS]; ++vmp) {
|
2010-03-29 13:39:54 +02:00
|
|
|
if (vmp->m_dev != NO_DEV && vmp->m_mounted_on) {
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
if (vmp->m_mounted_on->v_inode_nr == res.inode_nr &&
|
|
|
|
vmp->m_mounted_on->v_fs_e == res.fs_e) {
|
2007-08-07 14:52:47 +02:00
|
|
|
dir_vp = vmp->m_root_node;
|
|
|
|
break;
|
2009-09-21 16:49:26 +02:00
|
|
|
}
|
2007-08-07 14:52:47 +02:00
|
|
|
}
|
|
|
|
}
|
2012-02-13 16:28:04 +01:00
|
|
|
if (dir_vp == NULL) {
|
|
|
|
printf("VFS: path lookup error; root node not found\n");
|
|
|
|
if (vmpres) unlock_vmnt(vmpres);
|
|
|
|
*(resolve->l_vmp) = NULL;
|
|
|
|
return(EIO);
|
|
|
|
}
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
} else {
|
2007-08-07 14:52:47 +02:00
|
|
|
/* Climbing up mount */
|
|
|
|
/* Find the vmnt that represents the partition on
|
|
|
|
* which we "climb up". */
|
2010-05-10 15:26:00 +02:00
|
|
|
if ((vmp = find_vmnt(res.fs_e)) == NULL) {
|
2010-03-05 16:05:11 +01:00
|
|
|
panic("VFS lookup: can't find parent vmnt");
|
2012-02-13 16:28:04 +01:00
|
|
|
}
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
|
|
|
|
/* Make sure that the child FS does not feed a bogus path
|
|
|
|
* to the parent FS. That is, when we climb up the tree, we
|
|
|
|
* must've encountered ".." in the path, and that is exactly
|
|
|
|
* what we're going to feed to the parent */
|
2012-02-13 16:28:04 +01:00
|
|
|
if(strncmp(resolve->l_path, "..", 2) != 0 ||
|
|
|
|
(resolve->l_path[2] != '\0' && resolve->l_path[2] != '/')) {
|
|
|
|
printf("VFS: bogus path: %s\n", resolve->l_path);
|
|
|
|
if (vmpres) unlock_vmnt(vmpres);
|
|
|
|
*(resolve->l_vmp) = NULL;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(ENOENT);
|
|
|
|
}
|
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
/* Start node is the vnode on which the partition is
|
|
|
|
* mounted */
|
|
|
|
dir_vp = vmp->m_mounted_on;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set the starting directories inode number and FS endpoint */
|
|
|
|
fs_e = dir_vp->v_fs_e;
|
|
|
|
dir_ino = dir_vp->v_inode_nr;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
|
2007-08-07 14:52:47 +02:00
|
|
|
/* Is the process' root directory on the same partition?,
|
|
|
|
* if so, set the chroot directory too. */
|
2012-02-13 16:28:04 +01:00
|
|
|
if (dir_vp->v_dev == rfp->fp_rd->v_dev)
|
|
|
|
root_ino = rfp->fp_rd->v_inode_nr;
|
2007-08-07 14:52:47 +02:00
|
|
|
else
|
|
|
|
root_ino = 0;
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Unlock a previously locked vmnt if locked and lock new vmnt */
|
|
|
|
if (vmpres) unlock_vmnt(vmpres);
|
|
|
|
vmpres = find_vmnt(fs_e);
|
|
|
|
if (vmpres == NULL) return(EIO); /* mount point vanished? */
|
|
|
|
if ((r = lock_vmnt(vmpres, resolve->l_vmnt_lock)) != OK) {
|
|
|
|
if (r == EBUSY)
|
|
|
|
vmpres = NULL; /* Already locked */
|
|
|
|
else
|
|
|
|
return(r);
|
|
|
|
}
|
|
|
|
*(resolve->l_vmp) = vmpres;
|
|
|
|
|
|
|
|
r = req_lookup(fs_e, dir_ino, root_ino, uid, gid, resolve, &res, rfp);
|
2007-08-07 14:52:47 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (r != OK && r != EENTERMOUNT && r != ELEAVEMOUNT && r != ESYMLINK) {
|
|
|
|
if (vmpres) unlock_vmnt(vmpres);
|
|
|
|
*(resolve->l_vmp) = NULL;
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(r);
|
2012-02-13 16:28:04 +01:00
|
|
|
}
|
2007-01-05 17:36:55 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Fill in response fields */
|
2012-02-13 16:28:04 +01:00
|
|
|
result_node->inode_nr = res.inode_nr;
|
|
|
|
result_node->fmode = res.fmode;
|
|
|
|
result_node->fsize = res.fsize;
|
|
|
|
result_node->dev = res.dev;
|
|
|
|
result_node->fs_e = res.fs_e;
|
|
|
|
result_node->uid = res.uid;
|
|
|
|
result_node->gid = res.gid;
|
|
|
|
|
- Introduce support for sticky bit.
- Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly.
- Clean up MFS by removing old, dead code (backwards compatibility is broken by
the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all
functions have proper banners and prototypes.
- VFS should always provide a (syntactically) valid path to the FS; no need for
the FS to do sanity checks when leaving/entering mount points.
- Fix several bugs in MFS:
- Several path lookup bugs in MFS.
- A link can be too big for the path buffer.
- A mountpoint can become inaccessible when the creation of a new inode
fails, because the inode already exists and is a mountpoint.
- Introduce support for supplemental groups.
- Add test 46 to test supplemental group functionality (and removed obsolete
suppl. tests from test 2).
- Clean up VFS (not everything is done yet).
- ISOFS now opens device read-only. This makes the -r flag in the mount command
unnecessary (but will still report to be mounted read-write).
- Introduce PipeFS. PipeFS is a new FS that handles all anonymous and
named pipes. However, named pipes still reside on the (M)FS, as they are part
of the file system on disk. To make this work VFS now has a concept of
'mapped' inodes, which causes read, write, truncate and stat requests to be
redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
|
|
|
return(r);
|
2006-10-25 15:40:36 +02:00
|
|
|
}
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/*===========================================================================*
|
|
|
|
* lookup_init *
|
|
|
|
*===========================================================================*/
|
|
|
|
PUBLIC void lookup_init(resolve, path, flags, vmp, vp)
|
|
|
|
struct lookup *resolve;
|
|
|
|
char *path;
|
|
|
|
int flags;
|
|
|
|
struct vmnt **vmp;
|
|
|
|
struct vnode **vp;
|
|
|
|
{
|
|
|
|
assert(vmp != NULL);
|
|
|
|
assert(vp != NULL);
|
|
|
|
|
|
|
|
resolve->l_path = path;
|
|
|
|
resolve->l_flags = flags;
|
|
|
|
resolve->l_vmp = vmp;
|
|
|
|
resolve->l_vnode = vp;
|
|
|
|
resolve->l_vmnt_lock = TLL_NONE;
|
|
|
|
resolve->l_vnode_lock = TLL_NONE;
|
|
|
|
*vmp = NULL; /* Initialize lookup result to NULL */
|
|
|
|
*vp = NULL;
|
|
|
|
}
|
|
|
|
|
2010-08-30 15:44:07 +02:00
|
|
|
/*===========================================================================*
|
|
|
|
* get_name *
|
|
|
|
*===========================================================================*/
|
|
|
|
PUBLIC int get_name(dirp, entry, ename)
|
|
|
|
struct vnode *dirp;
|
|
|
|
struct vnode *entry;
|
|
|
|
char ename[NAME_MAX + 1];
|
|
|
|
{
|
2010-11-12 19:38:10 +01:00
|
|
|
u64_t pos, new_pos;
|
2010-08-30 15:44:07 +02:00
|
|
|
int r, consumed, totalbytes;
|
|
|
|
char buf[(sizeof(struct dirent) + NAME_MAX) * 8];
|
|
|
|
struct dirent *cur;
|
|
|
|
|
2010-11-12 19:38:10 +01:00
|
|
|
pos = make64(0, 0);
|
|
|
|
|
2010-08-30 15:44:07 +02:00
|
|
|
if ((dirp->v_mode & I_TYPE) != I_DIRECTORY) {
|
|
|
|
return(EBADF);
|
|
|
|
}
|
|
|
|
|
|
|
|
do {
|
2012-02-13 16:28:04 +01:00
|
|
|
r = req_getdents(dirp->v_fs_e, dirp->v_inode_nr, pos, buf, sizeof(buf),
|
|
|
|
&new_pos, 1);
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
if (r == 0) {
|
|
|
|
return(ENOENT); /* end of entries -- matching inode !found */
|
|
|
|
} else if (r < 0) {
|
|
|
|
return(r); /* error */
|
|
|
|
}
|
|
|
|
|
|
|
|
consumed = 0; /* bytes consumed */
|
|
|
|
totalbytes = r; /* number of bytes to consume */
|
|
|
|
|
|
|
|
do {
|
|
|
|
cur = (struct dirent *) (buf + consumed);
|
|
|
|
if (entry->v_inode_nr == cur->d_ino) {
|
|
|
|
/* found the entry we were looking for */
|
|
|
|
strncpy(ename, cur->d_name, NAME_MAX);
|
|
|
|
ename[NAME_MAX] = '\0';
|
|
|
|
return(OK);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* not a match -- move on to the next dirent */
|
|
|
|
consumed += cur->d_reclen;
|
|
|
|
} while (consumed < totalbytes);
|
|
|
|
|
|
|
|
pos = new_pos;
|
|
|
|
} while (1);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*===========================================================================*
|
|
|
|
* canonical_path *
|
|
|
|
*===========================================================================*/
|
2012-02-13 16:28:04 +01:00
|
|
|
PUBLIC int canonical_path(orig_path, rfp)
|
|
|
|
char orig_path[PATH_MAX];
|
2010-08-30 15:44:07 +02:00
|
|
|
struct fproc *rfp;
|
|
|
|
{
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Find canonical path of a given path */
|
2010-08-30 15:44:07 +02:00
|
|
|
int len = 0;
|
|
|
|
int r, symloop = 0;
|
|
|
|
struct vnode *dir_vp, *parent_dir;
|
2012-02-13 16:28:04 +01:00
|
|
|
struct vmnt *dir_vmp, *parent_vmp;
|
|
|
|
char component[NAME_MAX+1]; /* NAME_MAX does /not/ include '\0' */
|
|
|
|
char temp_path[PATH_MAX];
|
|
|
|
struct lookup resolve;
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
dir_vp = NULL;
|
2012-02-13 16:28:04 +01:00
|
|
|
strncpy(temp_path, orig_path, PATH_MAX);
|
|
|
|
temp_path[PATH_MAX - 1] = '\0';
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* First resolve path to the last directory holding the file */
|
2010-08-30 15:44:07 +02:00
|
|
|
do {
|
2012-02-13 16:28:04 +01:00
|
|
|
if (dir_vp) {
|
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
|
|
|
put_vnode(dir_vp);
|
2010-08-30 15:44:07 +02:00
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
lookup_init(&resolve, temp_path, PATH_NOFLAGS, &dir_vmp, &dir_vp);
|
|
|
|
resolve.l_vmnt_lock = VMNT_READ;
|
|
|
|
resolve.l_vnode_lock = VNODE_READ;
|
|
|
|
if ((dir_vp = last_dir(&resolve, rfp)) == NULL) return(err_code);
|
|
|
|
|
|
|
|
/* dir_vp points to dir and resolve path now contains only the
|
2010-08-30 15:44:07 +02:00
|
|
|
* filename.
|
|
|
|
*/
|
2012-02-13 16:28:04 +01:00
|
|
|
strncpy(orig_path, temp_path, NAME_MAX); /* Store file name */
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
/* check if the file is a symlink, if so resolve it */
|
2012-02-13 16:28:04 +01:00
|
|
|
r = rdlink_direct(orig_path, temp_path, rfp);
|
|
|
|
|
|
|
|
if (r <= 0)
|
2010-08-30 15:44:07 +02:00
|
|
|
break;
|
|
|
|
|
|
|
|
/* encountered a symlink -- loop again */
|
2012-02-13 16:28:04 +01:00
|
|
|
strncpy(orig_path, temp_path, PATH_MAX - 1);
|
2010-08-30 15:44:07 +02:00
|
|
|
symloop++;
|
|
|
|
} while (symloop < SYMLOOP_MAX);
|
|
|
|
|
|
|
|
if (symloop >= SYMLOOP_MAX) {
|
2012-02-13 16:28:04 +01:00
|
|
|
if (dir_vp) {
|
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
|
|
|
put_vnode(dir_vp);
|
|
|
|
}
|
|
|
|
return(ELOOP);
|
2010-08-30 15:44:07 +02:00
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* We've got the filename and the actual directory holding the file. From
|
|
|
|
* here we start building up the canonical path by climbing up the tree */
|
|
|
|
while (dir_vp != rfp->fp_rd) {
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
strcpy(temp_path, "..");
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
/* check if we're at the root node of the file system */
|
|
|
|
if (dir_vp->v_vmnt->m_root_node == dir_vp) {
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(dir_vp);
|
|
|
|
dir_vp = dir_vp->v_vmnt->m_mounted_on;
|
2012-02-13 16:28:04 +01:00
|
|
|
dir_vmp = dir_vp->v_vmnt;
|
|
|
|
if (lock_vmnt(dir_vmp, VMNT_READ) != OK)
|
|
|
|
panic("failed to lock vmnt");
|
|
|
|
if (lock_vnode(dir_vp, VNODE_READ) != OK)
|
|
|
|
panic("failed to lock vnode");
|
2010-08-30 15:44:07 +02:00
|
|
|
dup_vnode(dir_vp);
|
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
lookup_init(&resolve, temp_path, PATH_NOFLAGS, &parent_vmp,
|
|
|
|
&parent_dir);
|
|
|
|
resolve.l_vmnt_lock = VMNT_READ;
|
|
|
|
resolve.l_vnode_lock = VNODE_READ;
|
|
|
|
|
|
|
|
if ((parent_dir = advance(dir_vp, &resolve, rfp)) == NULL) {
|
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(dir_vp);
|
|
|
|
return(err_code);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* now we have to retrieve the name of the parent directory */
|
|
|
|
if (get_name(parent_dir, dir_vp, component) != OK) {
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(parent_dir);
|
|
|
|
unlock_vmnt(parent_vmp);
|
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(parent_dir);
|
2012-02-13 16:28:04 +01:00
|
|
|
put_vnode(dir_vp);
|
2010-08-30 15:44:07 +02:00
|
|
|
return(ENOENT);
|
|
|
|
}
|
|
|
|
|
|
|
|
len += strlen(component) + 1;
|
2011-09-12 11:00:24 +02:00
|
|
|
if (len >= PATH_MAX) {
|
2012-02-13 16:28:04 +01:00
|
|
|
/* adding the component to orig_path would exceed PATH_MAX */
|
|
|
|
unlock_vnode(parent_dir);
|
|
|
|
unlock_vmnt(parent_vmp);
|
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(parent_dir);
|
2012-02-13 16:28:04 +01:00
|
|
|
put_vnode(dir_vp);
|
2010-08-30 15:44:07 +02:00
|
|
|
return(ENOMEM);
|
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Store result of component in orig_path. First make space by moving
|
|
|
|
* the contents of orig_path to the right. Move strlen + 1 bytes to
|
|
|
|
* include the terminating '\0'. Move to strlen + 1 bytes to reserve
|
|
|
|
* space for the slash.
|
2010-08-30 15:44:07 +02:00
|
|
|
*/
|
2012-02-13 16:28:04 +01:00
|
|
|
memmove(orig_path+strlen(component)+1, orig_path, strlen(orig_path)+1);
|
2010-08-30 15:44:07 +02:00
|
|
|
/* Copy component into canon_path */
|
2012-02-13 16:28:04 +01:00
|
|
|
memmove(orig_path, component, strlen(component));
|
2010-08-30 15:44:07 +02:00
|
|
|
/* Put slash into place */
|
2012-02-13 16:28:04 +01:00
|
|
|
orig_path[strlen(component)] = '/';
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
/* Store parent_dir result, and continue the loop once more */
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(dir_vmp);
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(dir_vp);
|
|
|
|
dir_vp = parent_dir;
|
|
|
|
}
|
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(dir_vp);
|
|
|
|
unlock_vmnt(parent_vmp);
|
|
|
|
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(dir_vp);
|
|
|
|
|
|
|
|
/* add the leading slash */
|
2012-02-13 16:28:04 +01:00
|
|
|
if (strlen(orig_path) >= PATH_MAX) return(ENAMETOOLONG);
|
|
|
|
memmove(orig_path+1, orig_path, strlen(orig_path));
|
|
|
|
orig_path[0] = '/';
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
return(OK);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*===========================================================================*
|
|
|
|
* check_perms *
|
|
|
|
*===========================================================================*/
|
2012-02-13 16:28:04 +01:00
|
|
|
PRIVATE int check_perms(ep, io_gr, pathlen)
|
2010-08-30 15:44:07 +02:00
|
|
|
endpoint_t ep;
|
|
|
|
cp_grant_id_t io_gr;
|
2012-02-13 16:28:04 +01:00
|
|
|
size_t pathlen;
|
2010-08-30 15:44:07 +02:00
|
|
|
{
|
2012-02-13 16:28:04 +01:00
|
|
|
int r, slot;
|
2010-08-30 15:44:07 +02:00
|
|
|
struct vnode *vp;
|
2012-02-13 16:28:04 +01:00
|
|
|
struct vmnt *vmp;
|
2010-08-30 15:44:07 +02:00
|
|
|
struct fproc *rfp;
|
2011-09-12 11:00:24 +02:00
|
|
|
char canon_path[PATH_MAX];
|
2012-02-13 16:28:04 +01:00
|
|
|
struct lookup resolve;
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (isokendpt(ep, &slot) != OK) return(EINVAL);
|
|
|
|
if (pathlen < UNIX_PATH_MAX || pathlen >= PATH_MAX) return(EINVAL);
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
rfp = &(fproc[slot]);
|
2010-08-30 15:44:07 +02:00
|
|
|
r = sys_safecopyfrom(PFS_PROC_NR, io_gr, (vir_bytes) 0,
|
2012-02-13 16:28:04 +01:00
|
|
|
(vir_bytes) canon_path, pathlen, D);
|
|
|
|
if (r != OK) return(r);
|
|
|
|
canon_path[pathlen] = '\0';
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Turn path into canonical path to the socket file */
|
|
|
|
if ((r = canonical_path(canon_path, rfp)) != OK)
|
|
|
|
return(r);
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
if (strlen(canon_path) >= pathlen) return(ENAMETOOLONG);
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
/* copy canon_path back to PFS */
|
2012-02-13 16:28:04 +01:00
|
|
|
r = sys_safecopyto(PFS_PROC_NR, (cp_grant_id_t) io_gr, (vir_bytes) 0,
|
|
|
|
(vir_bytes) canon_path, pathlen, D);
|
|
|
|
if (r != OK) return(r);
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
/* Now do permissions checking */
|
|
|
|
lookup_init(&resolve, canon_path, PATH_NOFLAGS, &vmp, &vp);
|
|
|
|
resolve.l_vmnt_lock = VMNT_READ;
|
|
|
|
resolve.l_vnode_lock = VNODE_READ;
|
|
|
|
if ((vp = eat_path(&resolve, rfp)) == NULL) return(err_code);
|
2010-08-30 15:44:07 +02:00
|
|
|
|
|
|
|
/* check permissions */
|
2012-01-27 17:06:43 +01:00
|
|
|
r = forbidden(rfp, vp, (R_BIT | W_BIT));
|
2010-08-30 15:44:07 +02:00
|
|
|
|
2012-02-13 16:28:04 +01:00
|
|
|
unlock_vnode(vp);
|
|
|
|
unlock_vmnt(vmp);
|
|
|
|
|
2010-08-30 15:44:07 +02:00
|
|
|
put_vnode(vp);
|
|
|
|
return(r);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*===========================================================================*
|
|
|
|
* do_check_perms *
|
|
|
|
*===========================================================================*/
|
|
|
|
PUBLIC int do_check_perms(void)
|
|
|
|
{
|
Server/driver protocols: no longer allow third-party copies.
Before safecopies, the IO_ENDPT and DL_ENDPT message fields were needed
to know which actual process to copy data from/to, as that process may
not always be the caller. Now that we have full safecopy support, these
fields have become useless for that purpose: the owner of the grant is
*always* the caller. Allowing the caller to supply another endpoint is
in fact dangerous, because the callee may then end up using a grant
from a third party. One could call this a variant of the confused
deputy problem.
From now on, safecopy calls should always use the caller's endpoint as
grant owner. This fully obsoletes the DL_ENDPT field in the
inet/ethernet protocol. IO_ENDPT has other uses besides identifying the
grant owner though. This patch renames IO_ENDPT to USER_ENDPT, not only
because that is a more fitting name (it should never be used for I/O
after all), but also in order to intentionally break any old system
source code outside the base system. If this patch breaks your code,
fixing it is fairly simple:
- DL_ENDPT should be replaced with m_source;
- IO_ENDPT should be replaced with m_source when used for safecopies;
- IO_ENDPT should be replaced with USER_ENDPT for any other use, e.g.
when setting REP_ENDPT, matching requests in CANCEL calls, getting
DEV_SELECT flags, and retrieving of the real user process's endpoint
in DEV_OPEN.
The changes in this patch are binary backward compatible.
2011-04-11 19:35:05 +02:00
|
|
|
return check_perms(m_in.USER_ENDPT, (cp_grant_id_t) m_in.IO_GRANT,
|
2012-02-13 16:28:04 +01:00
|
|
|
(size_t) m_in.COUNT);
|
2010-08-30 15:44:07 +02:00
|
|
|
}
|