d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
4dbc7c2826
freebsd-dev/sys/fs/pseudofs/pseudofs_vnops.c

839 lines
20 KiB
C
Raw Normal View History

Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/*-
* Copyright (c) 2001 Dag-Erling Co<EFBFBD>dan Sm<EFBFBD>rgrav
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer
* in this position and unchanged.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/ctype.h>
#include <sys/dirent.h>
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
#include <sys/fcntl.h>
For some reason, though the module builds just fine without <sys/lock.h>, LINT fails to build without it.
2001-06-11 15:04:48 +00:00
#include <sys/lock.h>
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
#include <sys/mount.h>
Blah, not my day. This file needs <sys/mutex.h> now.
2001-06-10 10:42:55 +00:00
#include <sys/mutex.h>
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/sbuf.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/vnode.h>
#include <fs/pseudofs/pseudofs.h>
#include <fs/pseudofs/pseudofs_internal.h>
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
#if 0
#define PFS_TRACE(foo) \
do { \
Fix an incorrect PFS_TRACE. Also, use __func__ instead of __FUNCTION__.
2001-12-09 00:28:12 +00:00
printf("pseudofs: %s(): line %d: ", __func__, __LINE__); \
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
printf foo ; \
printf("\n"); \
} while (0)
#define PFS_RETURN(err) \
do { \
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
printf("pseudofs: %s(): line %d: returning %d\n", \
Fix an incorrect PFS_TRACE. Also, use __func__ instead of __FUNCTION__.
2001-12-09 00:28:12 +00:00
__func__, __LINE__, err); \
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
return (err); \
} while (0)
#else
#define PFS_TRACE(foo) \
do { /* nothing */ } while (0)
#define PFS_RETURN(err) \
return (err)
#endif
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/*
* Returns non-zero if given file is visible to given process
*/
static int
pfs_visible(struct thread *td, struct pfs_node *pn, pid_t pid)
{
struct proc *proc;
int r;
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
PFS_TRACE(("%s (pid: %d, req: %d)",
pn->pn_name, pid, td->td_proc->p_pid));
Add a PFS_DISABLED flag; pfs_visible() automatically returns 0 if it is set on the node in question. Also add two API functions for setting and clearing this flag; setting it also reclaims all vnodes associated with the node.
2001-10-02 22:22:42 +00:00
if (pn->pn_flags & PFS_DISABLED)
PFS_RETURN (0);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
r = 1;
Add a PFS_DISABLED flag; pfs_visible() automatically returns 0 if it is set on the node in question. Also add two API functions for setting and clearing this flag; setting it also reclaims all vnodes associated with the node.
2001-10-02 22:22:42 +00:00
if (pid != NO_PID) {
if ((proc = pfind(pid)) == NULL)
PFS_RETURN (0);
Change p_can{debug,see,sched,signal}()'s first argument to be a thread pointer instead of a proc pointer and require the process pointed to by the second argument to be locked. We now use the thread ucred reference for the credential checks in p_can*() as a result. p_canfoo() should now no longer need Giant.
2002-05-19 00:14:50 +00:00
if (p_cansee(td, proc) != 0 ||
Add a PFS_DISABLED flag; pfs_visible() automatically returns 0 if it is set on the node in question. Also add two API functions for setting and clearing this flag; setting it also reclaims all vnodes associated with the node.
2001-10-02 22:22:42 +00:00
(pn->pn_vis != NULL && !(pn->pn_vis)(td, proc, pn)))
r = 0;
PROC_UNLOCK(proc);
}
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
PFS_RETURN (r);
}
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/*
* Verify permissions
*/
static int
pfs_access(struct vop_access_args *va)
{
struct vnode *vn = va->a_vp;
struct vattr vattr;
int error;
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
PFS_TRACE((((struct pfs_vdata *)vn->v_data)->pvd_pn->pn_name));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
error = VOP_GETATTR(vn, &vattr, va->a_cred, va->a_td);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (error)
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
PFS_RETURN (error);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
error = vaccess(vn->v_type, vattr.va_mode, vattr.va_uid,
vattr.va_gid, va->a_mode, va->a_cred, NULL);
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
PFS_RETURN (error);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
* Close a file or directory
*/
static int
pfs_close(struct vop_close_args *va)
{
Add support for a last-close handler. Revert the module version bumps; they're quite pointless as long as the only pseudofs consumer is linprocfs, which is in the tree.
2001-11-27 13:26:27 +00:00
struct vnode *vn = va->a_vp;
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
struct proc *proc;
int error;
PFS_TRACE((pn->pn_name));
/*
* Do nothing unless this is the last close and the node has a
* last-close handler.
*/
- Use vrefcnt() where it is safe to do so instead of doing direct and unlocked accesses to v_usecount. - Lock access to the buf lists in the various sync routines. interlock locking could be avoided almost entirely in leaf filesystems if the fsync function had a generic helper.
2002-09-25 02:32:42 +00:00
if (vrefcnt(vn) > 1 || pn->pn_close == NULL)
Add support for a last-close handler. Revert the module version bumps; they're quite pointless as long as the only pseudofs consumer is linprocfs, which is in the tree.
2001-11-27 13:26:27 +00:00
PFS_RETURN (0);
if (pvd->pvd_pid != NO_PID)
proc = pfind(pvd->pvd_pid);
else
proc = NULL;
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add support for a last-close handler. Revert the module version bumps; they're quite pointless as long as the only pseudofs consumer is linprocfs, which is in the tree.
2001-11-27 13:26:27 +00:00
error = (pn->pn_close)(va->a_td, proc, pn);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add support for a last-close handler. Revert the module version bumps; they're quite pointless as long as the only pseudofs consumer is linprocfs, which is in the tree.
2001-11-27 13:26:27 +00:00
if (proc != NULL)
PROC_UNLOCK(proc);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add support for a last-close handler. Revert the module version bumps; they're quite pointless as long as the only pseudofs consumer is linprocfs, which is in the tree.
2001-11-27 13:26:27 +00:00
PFS_RETURN (error);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
* Get file attributes
*/
static int
pfs_getattr(struct vop_getattr_args *va)
{
struct vnode *vn = va->a_vp;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
struct vattr *vap = va->a_vap;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
struct proc *proc;
int error = 0;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_TRACE((pn->pn_name));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
VATTR_NULL(vap);
vap->va_type = vn->v_type;
vap->va_fileid = pn->pn_fileno;
vap->va_flags = 0;
vap->va_blocksize = PAGE_SIZE;
vap->va_bytes = vap->va_size = 0;
vap->va_fsid = vn->v_mount->mnt_stat.f_fsid.val[0];
vap->va_nlink = 1;
nanotime(&vap->va_ctime);
vap->va_atime = vap->va_mtime = vap->va_ctime;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
switch (pn->pn_type) {
case pfstype_procdir:
case pfstype_root:
case pfstype_dir:
vap->va_mode = 0555;
break;
case pfstype_file:
case pfstype_symlink:
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
vap->va_mode = 0444;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
break;
default:
printf("shouldn't be here!\n");
vap->va_mode = 0;
break;
}
if (pvd->pvd_pid != NO_PID) {
if ((proc = pfind(pvd->pvd_pid)) == NULL)
PFS_RETURN (ENOENT);
vap->va_uid = proc->p_ucred->cr_ruid;
vap->va_gid = proc->p_ucred->cr_rgid;
if (pn->pn_attr != NULL)
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
error = (pn->pn_attr)(va->a_td, proc, pn, vap);
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PROC_UNLOCK(proc);
} else {
vap->va_uid = 0;
vap->va_gid = 0;
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_RETURN (error);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
/*
* Perform an ioctl
*/
static int
pfs_ioctl(struct vop_ioctl_args *va)
{
struct vnode *vn = va->a_vp;
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
struct proc *proc = NULL;
int error;
Fix an incorrect PFS_TRACE. Also, use __func__ instead of __FUNCTION__.
2001-12-09 00:28:12 +00:00
PFS_TRACE(("%s: %lx", pn->pn_name, va->a_command));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
if (vn->v_type != VREG)
PFS_RETURN (EINVAL);
if (pn->pn_ioctl == NULL)
PFS_RETURN (ENOTTY);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
/*
fix grammar in comment
2003-06-20 23:29:04 +00:00
* This is necessary because process' privileges may
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
* have changed since the open() call.
*/
if (!pfs_visible(curthread, pn, pvd->pvd_pid))
PFS_RETURN (EIO);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
/* XXX duplicates bits of pfs_visible() */
if (pvd->pvd_pid != NO_PID) {
if ((proc = pfind(pvd->pvd_pid)) == NULL)
PFS_RETURN (EIO);
_PHOLD(proc);
PROC_UNLOCK(proc);
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
error = (pn->pn_ioctl)(curthread, proc, pn, va->a_command, va->a_data);
if (proc != NULL)
PRELE(proc);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
PFS_RETURN (error);
}
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
/*
* Perform getextattr
*/
static int
pfs_getextattr(struct vop_getextattr_args *va)
{
struct vnode *vn = va->a_vp;
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
struct proc *proc = NULL;
int error;
PFS_TRACE((pd->pn_name));
if (pn->pn_getextattr == NULL)
PFS_RETURN (EOPNOTSUPP);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
/*
* This is necessary because either process' privileges may
* have changed since the open() call.
*/
if (!pfs_visible(curthread, pn, pvd->pvd_pid))
PFS_RETURN (EIO);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
/* XXX duplicates bits of pfs_visible() */
if (pvd->pvd_pid != NO_PID) {
if ((proc = pfind(pvd->pvd_pid)) == NULL)
PFS_RETURN (EIO);
_PHOLD(proc);
PROC_UNLOCK(proc);
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
error = (pn->pn_getextattr)(curthread, proc, pn, va->a_attrnamespace,
Part I: Update extended attribute API and ABI: o Modify the system call syntax for extattr_{get,set}_{fd,file}() so as not to use the scatter gather API (which appeared not to be used by any consumers, and be less portable), rather, accepts 'data' and 'nbytes' in the style of other simple read/write interfaces. This changes the API and ABI. o Modify system call semantics so that extattr_get_{fd,file}() return a size_t. When performing a read, the number of bytes read will be returned, unless the data pointer is NULL, in which case the number of bytes of data are returned. This changes the API only. o Modify the VOP_GETEXTATTR() vnode operation to accept a *size_t argument so as to return the size, if desirable. If set to NULL, the size will not be returned. o Update various filesystems (pseodofs, ufs) to DTRT. These changes should make extended attributes more useful and more portable. More commits to rebuild the system call files, as well as update userland utilities to follow. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2002-02-10 04:43:22 +00:00
va->a_name, va->a_uio, va->a_size, va->a_cred);
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
if (proc != NULL)
PRELE(proc);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
PFS_RETURN (error);
}
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/*
* Look up a file or directory
Pseudofs was leaking VFS cache entries badly due to its cache and use of the wrong VOP descriptor. This misuse caused VFS-cached vnodes to be re-cached, resulting in the leak. This commit is an interim fix until DES has a chance to rework the code involved.
2001-12-19 23:58:09 +00:00
*
* XXX NOTE! pfs_lookup() has been hooked into vop_lookup_desc! This
* will result in a lookup operation for a vnode which may already be
* cached, therefore we have to be careful to purge the VFS cache when
* reusing a vnode.
*
* This code will work, but is not really correct. Normally we would hook
* vfs_cache_lookup() into vop_lookup_desc and hook pfs_lookup() into
* vop_cachedlookup_desc.
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
*/
static int
pfs_lookup(struct vop_lookup_args *va)
{
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct vnode *vn = va->a_dvp;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
struct vnode **vpp = va->a_vpp;
struct componentname *cnp = va->a_cnp;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pd = pvd->pvd_pn;
struct pfs_node *pn, *pdn = NULL;
pid_t pid = pvd->pvd_pid;
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
int lockparent;
int wantparent;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
char *pname;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
int error, i, namelen;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_TRACE(("%.*s", (int)cnp->cn_namelen, cnp->cn_nameptr));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
cnp->cn_flags &= ~PDIRUNLOCK;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (vn->v_type != VDIR)
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_RETURN (ENOTDIR);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
/*
* Don't support DELETE or RENAME. CREATE is supported so
* that O_CREAT will work, but the lookup will still fail if
* the file does not exist.
*/
if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EOPNOTSUPP);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
/* shortcut: check if the name is too long */
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (cnp->cn_namelen >= PFS_NAMELEN)
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_RETURN (ENOENT);
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/* check that parent directory is visisble... */
if (!pfs_visible(curthread, pd, pvd->pvd_pid))
PFS_RETURN (ENOENT);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
lockparent = cnp->cn_flags & LOCKPARENT;
wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* self */
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
namelen = cnp->cn_namelen;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
pname = cnp->cn_nameptr;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (namelen == 1 && *pname == '.') {
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
pn = pd;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
*vpp = vn;
VREF(vn);
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
PFS_RETURN (0);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/* parent */
if (cnp->cn_flags & ISDOTDOT) {
if (pd->pn_type == pfstype_root)
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_RETURN (EIO);
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
VOP_UNLOCK(vn, 0, cnp->cn_thread);
cnp->cn_flags |= PDIRUNLOCK;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
KASSERT(pd->pn_parent, ("non-root directory has no parent"));
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
/*
* This one is tricky. Descendents of procdir nodes
* inherit their parent's process affinity, but
* there's no easy reverse mapping. For simplicity,
* we assume that if this node is a procdir, its
* parent isn't (which is correct as long as
* descendents of procdir nodes are never procdir
* nodes themselves)
*/
if (pd->pn_type == pfstype_procdir)
pid = NO_PID;
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
pn = pd->pn_parent;
goto got_pnode;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
/* named node */
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
for (pn = pd->pn_nodes; pn != NULL; pn = pn->pn_next)
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (pn->pn_type == pfstype_procdir)
pdn = pn;
else if (pn->pn_name[namelen] == '\0'
&& bcmp(pname, pn->pn_name, namelen) == 0)
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
goto got_pnode;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
/* process dependent node */
if ((pn = pdn) != NULL) {
pid = 0;
for (pid = 0, i = 0; i < namelen && isdigit(pname[i]); ++i)
if ((pid = pid * 10 + pname[i] - '0') > PID_MAX)
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
break;
if (i == cnp->cn_namelen)
goto got_pnode;
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (ENOENT);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
got_pnode:
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
if (pn != pd->pn_parent && !pn->pn_parent)
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
pn->pn_parent = pd;
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
if (!pfs_visible(curthread, pn, pvd->pvd_pid))
PFS_RETURN (ENOENT);
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
error = pfs_vncache_alloc(vn->v_mount, vpp, pn, pid);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (error)
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_RETURN (error);
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
if ((cnp->cn_flags & ISDOTDOT) && (cnp->cn_flags & ISLASTCN)
&& lockparent) {
vn_lock(vn, LK_EXCLUSIVE|LK_RETRY, cnp->cn_thread);
cnp->cn_flags &= ~PDIRUNLOCK;
}
Don't unlock the parent directory vnode twice if the ISDOTDOT flag is set.
2003-06-01 09:16:26 +00:00
if (!((lockparent && (cnp->cn_flags & ISLASTCN)) ||
(cnp->cn_flags & ISDOTDOT)))
Lock down pseudofs: - Initialize lock structure in vncache_alloc - Return locked vnodes from vncache_alloc - Setup vnode op vectors to use default lock, unlock, and islocked - Implement simple locking scheme required for lookup
2002-07-08 01:50:14 +00:00
VOP_UNLOCK(vn, 0, cnp->cn_thread);
Pseudofs was leaking VFS cache entries badly due to its cache and use of the wrong VOP descriptor. This misuse caused VFS-cached vnodes to be re-cached, resulting in the leak. This commit is an interim fix until DES has a chance to rework the code involved.
2001-12-19 23:58:09 +00:00
/*
* XXX See comment at top of the routine.
*/
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (cnp->cn_flags & MAKEENTRY)
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
cache_enter(vn, *vpp, cnp);
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_RETURN (0);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
* Open a file or directory.
*/
static int
pfs_open(struct vop_open_args *va)
{
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
struct vnode *vn = va->a_vp;
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
int mode = va->a_mode;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_TRACE(("%s (mode 0x%x)", pn->pn_name, mode));
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/*
* check if the file is visible to the caller
*
* XXX Not sure if this is necessary, as the VFS system calls
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
* XXX pfs_lookup() and pfs_access() first, and pfs_lookup()
* XXX calls pfs_visible(). There's a race condition here, but
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
* XXX calling pfs_visible() from here doesn't really close it,
* XXX and the only consequence of that race is an EIO further
* XXX down the line.
*/
if (!pfs_visible(va->a_td, pn, pvd->pvd_pid))
PFS_RETURN (ENOENT);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
/* check if the requested mode is permitted */
if (((mode & FREAD) && !(mode & PFS_RD)) ||
((mode & FWRITE) && !(mode & PFS_WR)))
PFS_RETURN (EPERM);
/* we don't support locking */
if ((mode & O_SHLOCK) || (mode & O_EXLOCK))
PFS_RETURN (EOPNOTSUPP);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
PFS_RETURN (0);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
* Read from a file
*/
static int
pfs_read(struct vop_read_args *va)
{
struct vnode *vn = va->a_vp;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
struct uio *uio = va->a_uio;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct proc *proc = NULL;
Catch up with the change in sbuf_new's prototype.
2001-06-10 10:34:21 +00:00
struct sbuf *sb = NULL;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
char *ps;
int error, xlen;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_TRACE((pn->pn_name));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (vn->v_type != VREG)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EINVAL);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
if (!(pn->pn_flags & PFS_RD))
PFS_RETURN (EBADF);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
if (pn->pn_func == NULL)
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
PFS_RETURN (EIO);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/*
* This is necessary because either process' privileges may
* have changed since the open() call.
*/
if (!pfs_visible(curthread, pn, pvd->pvd_pid))
PFS_RETURN (EIO);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/* XXX duplicates bits of pfs_visible() */
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (pvd->pvd_pid != NO_PID) {
if ((proc = pfind(pvd->pvd_pid)) == NULL)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EIO);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
_PHOLD(proc);
PROC_UNLOCK(proc);
}
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
if (pn->pn_flags & PFS_RAWRD) {
The previous commit introduced some references to "curproc" which should have been references to "curthread". Correct this.
2001-09-28 12:36:54 +00:00
error = (pn->pn_func)(curthread, proc, pn, NULL, uio);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
if (proc != NULL)
PRELE(proc);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Catch up with the change in sbuf_new's prototype.
2001-06-10 10:34:21 +00:00
sb = sbuf_new(sb, NULL, uio->uio_offset + uio->uio_resid, 0);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (sb == NULL) {
if (proc != NULL)
PRELE(proc);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EIO);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
}
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
error = (pn->pn_func)(curthread, proc, pn, sb, uio);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (proc != NULL)
PRELE(proc);
Bail out if the fill function failed.
2001-06-10 21:39:01 +00:00
if (error) {
sbuf_delete(sb);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Bail out if the fill function failed.
2001-06-10 21:39:01 +00:00
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* XXX we should possibly detect and handle overflows */
Catch up with the change in sbuf_new's prototype.
2001-06-10 10:34:21 +00:00
sbuf_finish(sb);
ps = sbuf_data(sb) + uio->uio_offset;
xlen = sbuf_len(sb) - uio->uio_offset;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
xlen = imin(xlen, uio->uio_resid);
error = (xlen <= 0 ? 0 : uiomove(ps, xlen, uio));
Catch up with the change in sbuf_new's prototype.
2001-06-10 10:34:21 +00:00
sbuf_delete(sb);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
/*
* Iterate through directory entries
*/
static int
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
pfs_iterate(struct thread *td, pid_t pid, struct pfs_node *pd,
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
struct pfs_node **pn, struct proc **p)
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
{
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
if ((*pn) == NULL)
*pn = pd->pn_nodes;
else
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
again:
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if ((*pn)->pn_type != pfstype_procdir)
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
*pn = (*pn)->pn_next;
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
while (*pn != NULL && (*pn)->pn_type == pfstype_procdir) {
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (*p == NULL)
*p = LIST_FIRST(&allproc);
else
*p = LIST_NEXT(*p, p_list);
if (*p != NULL)
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
break;
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
*pn = (*pn)->pn_next;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
if ((*pn) == NULL)
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
return (-1);
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
if (!pfs_visible(td, *pn, *p ? (*p)->p_pid : pid))
goto again;
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
return (0);
}
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/*
* Return directory entries.
*/
static int
pfs_readdir(struct vop_readdir_args *va)
{
struct vnode *vn = va->a_vp;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct pfs_info *pi = (struct pfs_info *)vn->v_mount->mnt_data;
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pd = pvd->pvd_pn;
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
pid_t pid = pvd->pvd_pid;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct pfs_node *pn;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
struct dirent entry;
struct uio *uio;
struct proc *p;
off_t offset;
int error, i, resid;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_TRACE((pd->pn_name));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (vn->v_type != VDIR)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (ENOTDIR);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
uio = va->a_uio;
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/* check if the directory is visible to the caller */
if (!pfs_visible(curthread, pd, pid))
PFS_RETURN (ENOENT);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* only allow reading entire entries */
offset = uio->uio_offset;
resid = uio->uio_resid;
if (offset < 0 || offset % PFS_DELEN != 0 || resid < PFS_DELEN)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EINVAL);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* skip unwanted entries */
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
sx_slock(&allproc_lock);
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
for (pn = NULL, p = NULL; offset > 0; offset -= PFS_DELEN)
if (pfs_iterate(curthread, pid, pd, &pn, &p) == -1) {
/* nothing left... */
sx_sunlock(&allproc_lock);
PFS_RETURN (0);
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* fill in entries */
entry.d_reclen = PFS_DELEN;
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
while (pfs_iterate(curthread, pid, pd, &pn, &p) != -1 && resid > 0) {
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (!pn->pn_parent)
pn->pn_parent = pd;
if (!pn->pn_fileno)
pfs_fileno_alloc(pi, pn);
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
if (pid != NO_PID)
entry.d_fileno = pn->pn_fileno * NO_PID + pid;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
else
entry.d_fileno = pn->pn_fileno;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* PFS_DELEN was picked to fit PFS_NAMLEN */
for (i = 0; i < PFS_NAMELEN - 1 && pn->pn_name[i] != '\0'; ++i)
entry.d_name[i] = pn->pn_name[i];
entry.d_name[i] = 0;
entry.d_namlen = i;
switch (pn->pn_type) {
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
case pfstype_procdir:
KASSERT(p != NULL,
("reached procdir node with p == NULL"));
entry.d_fileno = pn->pn_fileno * NO_PID + p->p_pid;
entry.d_namlen = snprintf(entry.d_name,
PFS_NAMELEN, "%d", p->p_pid);
/* fall through */
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
case pfstype_root:
case pfstype_dir:
case pfstype_this:
case pfstype_parent:
entry.d_type = DT_DIR;
break;
case pfstype_file:
entry.d_type = DT_REG;
break;
case pfstype_symlink:
entry.d_type = DT_LNK;
break;
default:
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
sx_sunlock(&allproc_lock);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
panic("%s has unexpected node type: %d", pn->pn_name, pn->pn_type);
}
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
PFS_TRACE((entry.d_name));
Get rid of caddr_t.
2003-03-02 22:23:45 +00:00
if ((error = uiomove(&entry, PFS_DELEN, uio))) {
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
sx_sunlock(&allproc_lock);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
}
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
offset += PFS_DELEN;
resid -= PFS_DELEN;
}
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
sx_sunlock(&allproc_lock);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
uio->uio_offset += offset;
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (0);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
* Read a symbolic link
*/
static int
pfs_readlink(struct vop_readlink_args *va)
{
struct vnode *vn = va->a_vp;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
struct uio *uio = va->a_uio;
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
struct proc *proc = NULL;
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
char buf[MAXPATHLEN], *ps;
struct sbuf sb;
int error, xlen;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_TRACE((pn->pn_name));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
if (vn->v_type != VLNK)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EINVAL);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
if (pn->pn_func == NULL)
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
PFS_RETURN (EIO);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (pvd->pvd_pid != NO_PID) {
if ((proc = pfind(pvd->pvd_pid)) == NULL)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EIO);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
_PHOLD(proc);
PROC_UNLOCK(proc);
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* sbuf_new() can't fail with a static buffer */
sbuf_new(&sb, buf, sizeof buf, 0);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
error = (pn->pn_func)(curthread, proc, pn, &sb, NULL);
Add support for process-dependent directories. This means that save for the lack of a man page, pseudofs is mostly complete now.
2001-06-10 18:39:21 +00:00
if (proc != NULL)
PRELE(proc);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Bail out if the fill function failed.
2001-06-10 21:39:01 +00:00
if (error) {
sbuf_delete(&sb);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Bail out if the fill function failed.
2001-06-10 21:39:01 +00:00
}
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/* XXX we should detect and handle overflows */
sbuf_finish(&sb);
ps = sbuf_data(&sb) + uio->uio_offset;
xlen = sbuf_len(&sb) - uio->uio_offset;
xlen = imin(xlen, uio->uio_resid);
error = (xlen <= 0 ? 0 : uiomove(ps, xlen, uio));
sbuf_delete(&sb);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
* Reclaim a vnode
*/
static int
pfs_reclaim(struct vop_reclaim_args *va)
{
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
PFS_TRACE((((struct pfs_vdata *)va->a_vp->v_data)->pvd_pn->pn_name));
Pseudofs was leaking VFS cache entries badly due to its cache and use of the wrong VOP descriptor. This misuse caused VFS-cached vnodes to be re-cached, resulting in the leak. This commit is an interim fix until DES has a chance to rework the code involved.
2001-12-19 23:58:09 +00:00
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
return (pfs_vncache_free(va->a_vp));
}
/*
* Set attributes
*/
static int
pfs_setattr(struct vop_setattr_args *va)
{
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
PFS_TRACE((((struct pfs_vdata *)va->a_vp->v_data)->pvd_pn->pn_name));
Change EPERM to EOPNOTSUPP when failing pseudofs_setattr() arbitrarily. Quoth the alfred: The latter would be better.
2002-02-04 18:21:59 +00:00
PFS_RETURN (EOPNOTSUPP);
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
}
/*
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
* Read from a file
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
*/
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
static int
pfs_write(struct vop_read_args *va)
{
struct vnode *vn = va->a_vp;
struct pfs_vdata *pvd = (struct pfs_vdata *)vn->v_data;
struct pfs_node *pn = pvd->pvd_pn;
struct uio *uio = va->a_uio;
struct proc *proc = NULL;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
struct sbuf sb;
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
int error;
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
PFS_TRACE((pn->pn_name));
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
if (vn->v_type != VREG)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EINVAL);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
if (!(pn->pn_flags & PFS_WR))
PFS_RETURN (EBADF);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
Switch to dynamic rather than static initialization. This makes it possible (in theory) for nodes to be added and / or removed from pseudofs filesystems at runtime.
2001-10-19 01:43:06 +00:00
if (pn->pn_func == NULL)
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
PFS_RETURN (EIO);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
YA pseudofs megacommit, part 1: - Add a third callback to the pfs_node structure. This one simply returns non-zero if the specified requesting process is allowed to access the specified node for the specified target process. This is used in addition to the usual permission checks, e.g. when certain files don't make sense for certain (system) processes. - Make sure that pfs_lookup() and pfs_readdir() don't yap about files which aren't pfs_visible(). Also check pfs_visible() before performing reads and writes, to prevent the kind of races reported in SA-00:77 and SA-01:55 (fork a child, open /proc/child/ctl, have that child fork a setuid binary, and assume control of it). - Add some more trace points.
2001-10-01 04:22:20 +00:00
/*
* This is necessary because either process' privileges may
* have changed since the open() call.
*/
if (!pfs_visible(curthread, pn, pvd->pvd_pid))
PFS_RETURN (EIO);
/* XXX duplicates bits of pfs_visible() */
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
if (pvd->pvd_pid != NO_PID) {
if ((proc = pfind(pvd->pvd_pid)) == NULL)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (EIO);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
_PHOLD(proc);
PROC_UNLOCK(proc);
}
if (pn->pn_flags & PFS_RAWWR) {
The previous commit introduced some references to "curproc" which should have been references to "curthread". Correct this.
2001-09-28 12:36:54 +00:00
error = (pn->pn_func)(curthread, proc, pn, NULL, uio);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
if (proc != NULL)
PRELE(proc);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
}
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
sbuf_uionew(&sb, uio, &error);
if (error)
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
error = (pn->pn_func)(curthread, proc, pn, &sb, uio);
if (proc != NULL)
PRELE(proc);
Gratuitous whitespace cleanup.
2002-06-06 16:59:24 +00:00
Pseudofs take 2: - Remove hardcoded uid, gid, mode from struct pfs_node; make pfs_getattr() smart enough to get it right most of the time, and allow for callbacks to handle the remaining cases. Rework the definition macros to match. - Add lots of (conditional) debugging output. - Fix a long-standing bug inherited from procfs: don't pretend to be a read-only file system. Instead, return EOPNOTSUPP for operations we truly can't support and allow others to fail silently. In particular, pfs_lookup() now treats CREATE as LOOKUP. This may need more work. - In pfs_lookup(), if the parent node is process-dependent, check that the process in question still exists. - Implement pfs_open() - its only current function is to check that the process opening the file can see the process it belongs to. - Finish adding support for writeable nodes. - Bump module version number. - Introduce lots of new bugs.
2001-09-29 00:49:29 +00:00
sbuf_delete(&sb);
pseudofs.h: - Rearrange the flag constants a little to simplify specifying and testing for readability and writeability. pseudofs_vnops.c: - Track the aforementioned change. - Add checks to pfs_open() to prevent opening read-only files for writing or vice versa (pfs_{read,write} would block the actual reads and writes, but it's still a bug to allow the open() to succeed). Also, return EOPNOTSUPP if the caller attempts to lock the file. - Add more trace points.
2001-09-30 19:41:29 +00:00
PFS_RETURN (error);
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
}
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
/*
* Vnode operations
*/
vop_t **pfs_vnodeop_p;
static struct vnodeopv_entry_desc pfs_vnodeop_entries[] = {
{ &vop_default_desc, (vop_t *)vop_defaultop },
{ &vop_access_desc, (vop_t *)pfs_access },
{ &vop_close_desc, (vop_t *)pfs_close },
s/pfs_badop/vop_eopnotsupp/ Submitted by: phk
2002-05-03 14:58:25 +00:00
{ &vop_create_desc, (vop_t *)vop_eopnotsupp },
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
{ &vop_getattr_desc, (vop_t *)pfs_getattr },
Add VOP_GETEXTATTR(9) passthrough support to pseudofs. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2001-12-11 20:48:20 +00:00
{ &vop_getextattr_desc, (vop_t *)pfs_getextattr },
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or symlink lacked a filler function.
2001-10-26 18:52:47 +00:00
{ &vop_ioctl_desc, (vop_t *)pfs_ioctl },
s/pfs_badop/vop_eopnotsupp/ Submitted by: phk
2002-05-03 14:58:25 +00:00
{ &vop_link_desc, (vop_t *)vop_eopnotsupp },
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
{ &vop_lookup_desc, (vop_t *)pfs_lookup },
s/pfs_badop/vop_eopnotsupp/ Submitted by: phk
2002-05-03 14:58:25 +00:00
{ &vop_mkdir_desc, (vop_t *)vop_eopnotsupp },
{ &vop_mknod_desc, (vop_t *)vop_eopnotsupp },
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
{ &vop_open_desc, (vop_t *)pfs_open },
{ &vop_read_desc, (vop_t *)pfs_read },
{ &vop_readdir_desc, (vop_t *)pfs_readdir },
{ &vop_readlink_desc, (vop_t *)pfs_readlink },
{ &vop_reclaim_desc, (vop_t *)pfs_reclaim },
s/pfs_badop/vop_eopnotsupp/ Submitted by: phk
2002-05-03 14:58:25 +00:00
{ &vop_remove_desc, (vop_t *)vop_eopnotsupp },
{ &vop_rename_desc, (vop_t *)vop_eopnotsupp },
{ &vop_rmdir_desc, (vop_t *)vop_eopnotsupp },
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
{ &vop_setattr_desc, (vop_t *)pfs_setattr },
s/pfs_badop/vop_eopnotsupp/ Submitted by: phk
2002-05-03 14:58:25 +00:00
{ &vop_symlink_desc, (vop_t *)vop_eopnotsupp },
Clean up my source tree to avoid getting hit too badly by the next KSE or whatever mega-commit. This goes some way towards adding support for writeable files (needed by procfs).
2001-09-25 13:25:30 +00:00
{ &vop_write_desc, (vop_t *)pfs_write },
s/pfs_badop/vop_eopnotsupp/ Submitted by: phk
2002-05-03 14:58:25 +00:00
/* XXX I've probably forgotten a few that need vop_eopnotsupp */
Let pseudofs into the warmth of the FreeBSD CVS repo. It's not finished yet (I still have to find a way to implement process- dependent nodes without consuming too much memory, and the permission system needs tightening up), but it's becoming hard to work on without a repo (I've accidentally almost nuked it once already), and it works (except for the lack of process-dependent nodes, that is). I was supposed to commit this a week ago, but timed out waiting for jkh to reply to some questions I had. Pass him a spoonful of bad karma :)
2001-04-07 19:51:12 +00:00
{ NULL, (vop_t *)NULL }
};
static struct vnodeopv_desc pfs_vnodeop_opv_desc =
{ &pfs_vnodeop_p, pfs_vnodeop_entries };
VNODEOP_SET(pfs_vnodeop_opv_desc);
Reference in New Issue Copy Permalink