Overhaul the audit filename handling
Subsystems (3)
cover Overhaul the audit filename handling
This is the same patchset that was sent earlier with the RFC prefix, the only differences are the inclusion of a reviewer and the fact that I've stripped the RFC prefix. I'm including this in the audit-next branch and plan on pushing this to Linus in the next merge window.
For those of you who didn't bother looking at the RFC posting, this patchset resolves various problems in the way audit handles filenames. The patches rework the way audit handles filenames, removing a lot of nasty hacks added recently, and fixing a few bugs that still remain. Most significant to folks outside of audit, patch 5/5 does away with the nasty getname()/putname() kludge in favor of a less ugly reference count approach.
---
Paul Moore (5):
fs: rework getname_kernel to handle up to PATH_MAX sized filenames
fs: create proper filename objects using getname_kernel()
audit: enable filename recording via getname_kernel()
audit: fix filename matching in __audit_inode() and __audit_inode_child()
audit: replace getname()/putname() hacks with reference counters
fs/exec.c | 11 +++ fs/namei.c | 98 ++++++++++++++++++----------- fs/open.c | 11 +++ include/linux/audit.h | 3 - include/linux/fs.h | 9 +-- kernel/audit.h | 17 +---- kernel/auditsc.c | 167 +++++++++---------------------------------------- 7 files changed, 115 insertions(+), 201 deletions(-)
PATCH 1/5 fs: rework getname_kernel to handle up to PATH_MAX sized filenames
In preparation for expanded use in the kernel, make getname_kernel() more useful by allowing it to handle any legal filename length.
--- fs/namei.c | 34 ++++++++++++++++++++-------------- 1 file changed, 20 insertions(+), 14 deletions(-)
fs/namei.c +20 -14
@@ -211,32 +211,38 @@ getname(const char __user * filename)return getname_flags (filename,0 ,NULL ); } -/* -* The "getname_kernel()" interface doesn't do pathnames longer -* than EMBEDDED_NAME_MAX. Deal with it - you're a kernel user. -*/ struct filename *getname_kernel (const char * filename) {struct filename *result; -char *kname; -int len; - - len =strlen (filename); -if (len >=EMBEDDED_NAME_MAX ) -return ERR_PTR (-ENAMETOOLONG ); +int len =strlen (filename) +1 ; result =__getname ();if (unlikely (!result))return ERR_PTR (-ENOMEM ); - kname = (char *)result +sizeof (*result); - result->name = kname; +if (len <=EMBEDDED_NAME_MAX ) { + result->name = (char *)(result) +sizeof (*result); + result->separate = false; + }else if (len <=PATH_MAX ) { +struct filename *tmp; + + tmp =kmalloc (sizeof (*tmp),GFP_KERNEL ); +if (unlikely (!tmp)) { +__putname (result); +return ERR_PTR (-ENOMEM ); + } + tmp->name = (char *)result; + tmp->separate = true; + result = tmp; + }else { +__putname (result); +return ERR_PTR (-ENAMETOOLONG ); + } +strlcpy ((char *)result->name, filename, len); result->uptr =NULL ; result->aname =NULL ; - result->separate = false; -strlcpy (kname, filename,EMBEDDED_NAME_MAX );return result; }
PATCH 2/5 fs: create proper filename objects using getname_kernel()
In reply to Paul MooreThere are several areas in the kernel that create temporary filename objects using the following pattern:
int func(const char *name)
{
struct filename *file = { .name = name };
...
return 0;
}... which for the most part works okay, but it causes havoc within the audit subsystem as the filename object does not persist beyond the lifetime of the function. This patch converts all of these temporary filename objects into proper filename objects using getname_kernel() and putname() which ensure that the filename object persists until the audit subsystem is finished with it.
--- fs/exec.c | 11 +++++++++-- fs/namei.c | 34 ++++++++++++++++++++++++++-------- fs/open.c | 11 +++++++++-- 3 files changed, 44 insertions(+), 12 deletions(-)
fs/exec.c +9 -2
@@ -789,8 +789,15 @@ exit:struct file *open_exec (const char *name) { -struct filename tmp = { .name = name }; -return do_open_exec (&tmp); +struct file *file; +struct filename *tmp; + + tmp =getname_kernel (name); +if (unlikely (IS_ERR (tmp))) +return (void *)tmp; + file =do_open_exec (tmp); +putname (tmp); +return file; }EXPORT_SYMBOL (open_exec);
fs/namei.c +26 -8
@@ -2001,9 +2001,15 @@ static int filename_lookup(int dfd, struct filename *name,static int do_path_lookup (int dfd,const char *name,unsigned int flags,struct nameidata *nd) { -struct filename filename = { .name = name }; +int retval; +struct filename *filename; -return filename_lookup (dfd, &filename, flags, nd); + filename =getname_kernel (name); +if (unlikely (IS_ERR (filename))) +return PTR_ERR (filename); + retval =filename_lookup (dfd, filename, flags, nd); +putname (filename); +return retval; }/* does lookup, returns the object with parent locked */
@@ -2368,8 +2374,15 @@ int kern_path_mountpoint(int dfd, const char *name, struct path *path, unsigned int flags) { - struct filename s = {.name = name}; -return filename_mountpoint (dfd, &s, path, flags); + int retval; +struct filename *s; + + s =getname_kernel (name); +if (unlikely (IS_ERR (s))) +return PTR_ERR (s); + retval =filename_mountpoint (dfd, s, path, flags); +putname (s); +return retval; }EXPORT_SYMBOL (kern_path_mountpoint);
@@ -3259,7 +3272,7 @@ struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, {struct nameidata nd;struct file *file; -struct filename filename = { .name = name }; +struct filename *filename;int flags = op->lookup_flags |LOOKUP_ROOT ; nd.root.mnt = mnt;
@@ -3268,11 +3281,16 @@ struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt,if (d_is_symlink (dentry) && op->intent &LOOKUP_OPEN )return ERR_PTR (-ELOOP ); - file =path_openat (-1 , &filename, &nd, op, flags |LOOKUP_RCU ); + filename =getname_kernel (name); +if (unlikely (IS_ERR (filename))) +return (void *)filename; + + file =path_openat (-1 , filename, &nd, op, flags |LOOKUP_RCU );if (unlikely (file ==ERR_PTR (-ECHILD ))) - file =path_openat (-1 , &filename, &nd, op, flags); + file =path_openat (-1 , filename, &nd, op, flags);if (unlikely (file ==ERR_PTR (-ESTALE ))) - file =path_openat (-1 , &filename, &nd, op, flags |LOOKUP_REVAL ); + file =path_openat (-1 , filename, &nd, op, flags |LOOKUP_REVAL ); +putname (filename);return file; }
fs/open.c +9 -2
@@ -940,8 +940,15 @@ struct file *file_open_name(struct filename *name, int flags, umode_t mode)*/ struct file *filp_open (const char *filename,int flags,umode_t mode) { -struct filename name = {.name = filename}; -return file_open_name (&name, flags, mode); +struct file *file; +struct filename *name; + + name =getname_kernel (filename); +if (unlikely (IS_ERR (name))) +return (void *)name; + file =file_open_name (name, flags, mode); +putname (name); +return file; }EXPORT_SYMBOL (filp_open);
Re: [PATCH 2/5] fs: create proper filename objects using getname_kernel()
Quoted 10 lines · Paul Moore
There are several areas in the kernel that create temporary filename objects using the following pattern:
int func(const char *name) { struct filename *file = { .name = name }; ... return 0; }
Quoted 6 lines
... which for the most part works okay, but it causes havoc within the audit subsystem as the filename object does not persist beyond the lifetime of the function. This patch converts all of these temporary filename objects into proper filename objects using
getname_kernel()andputname()which ensure that the filename object persists until the audit subsystem is finished with it.
Hi Paul,
With this patch (bisected) my vm fails to boot with a virtio-9p rootfs:
[ 27.313687] fa00 2097152 vda driver: virtio_blk [ 27.314218] 103:00000 8192 sda driver: sd [ 27.314714] DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify explicit textual name for "root=" boot option. [ 27.315963] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0) [ 27.316885] CPU: 29 PID: 1 Comm: swapper/0 Not tainted 3.19.0-rc5-next-20150120-sasha-00053-gb2e3c55-dirty #1772 [ 27.316885] ffffffffffffffff ffff88005cb8bd28 ffffffffb14522b5 000000000000004e [ 27.316885] ffffffffb26d4950 ffff88005cb8bda8 ffffffffb144c298 ffff88005cb8bd48 [ 27.316885] ffffffff00000010 ffff88005cb8bdb8 ffff88005cb8bd58 fffffffffffffffe [ 27.316885] Call Trace: [ 27.316885] [<ffffffffb14522b5>] dump_stack+0x4f/0x7b [ 27.316885] [<ffffffffb144c298>] panic+0xd2/0x216 [ 27.316885] [<ffffffffb431558b>] mount_block_root+0x18b/0x244 [ 27.316885] [<ffffffffb43158ae>] mount_root+0x128/0x133 [ 27.316885] [<ffffffffb4315a1b>] prepare_namespace+0x162/0x19b [ 27.316885] [<ffffffffb4315223>] kernel_init_freeable+0x285/0x29a [ 27.316885] [<ffffffffb14472b0>] ? rest_init+0xd0/0xd0 [ 27.316885] [<ffffffffb14472be>] kernel_init+0xe/0xf0 [ 27.316885] [<ffffffffb148007c>] ret_from_fork+0x7c/0xb0 [ 27.316885] [<ffffffffb14472b0>] ? rest_init+0xd0/0xd0 [ 27.316885] Dumping ftrace buffer: [ 27.316885] (ftrace buffer empty) [ 27.316885] Kernel Offset: 0x2d000000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff) [ 27.316885] Rebooting in 1 seconds..
virtio-9p is not even listed as an option now.
Thanks,
Sasha
Re: [PATCH 2/5] fs: create proper filename objects using getname_kernel()
Quoted 48 lines · Sasha Levin
Quoted 12 lines · Paul Moore
There are several areas in the kernel that create temporary filename
objects using the following pattern:
int func(const char *name) { struct filename *file = { .name = name }; ... return 0; }Quoted 6 lines
... which for the most part works okay, but it causes havoc within the audit subsystem as the filename object does not persist beyond the lifetime of the function. This patch converts all of these temporary filename objects into proper filename objects using
getname_kernel()andputname()which ensure that the filename object persists until the audit subsystem is finished with it.Hi Paul,
With this patch (bisected) my vm fails to boot with a virtio-9p rootfs:
[ 27.313687] fa00 2097152 vda driver: virtio_blk [ 27.314218] 103:00000 8192 sda driver: sd [ 27.314714] DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify explicit textual name for "root=" boot option. [ 27.315963] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0) [ 27.316885] CPU: 29 PID: 1 Comm: swapper/0 Not tainted3.19.0-rc5-next-20150120-sasha-00053-gb2e3c55-dirty #1772 [ 27.316885]
ffffffffffffffffffff88005cb8bd28ffffffffb14522b5000000000000004e[ 27.316885]ffffffffb26d4950ffff88005cb8bda8ffffffffb144c298ffff88005cb8bd48[ 27.316885]ffffffff00000010ffff88005cb8bdb8ffff88005cb8bd58fffffffffffffffe [ 27.316885] Call Trace:[ 27.316885] [<ffffffffb14522b5>] dump_stack+0x4f/0x7b [ 27.316885] [<ffffffffb144c298>] panic+0xd2/0x216 [ 27.316885] [<ffffffffb431558b>] mount_block_root+0x18b/0x244 [ 27.316885] [<ffffffffb43158ae>] mount_root+0x128/0x133 [ 27.316885] [<ffffffffb4315a1b>] prepare_namespace+0x162/0x19b [ 27.316885] [<ffffffffb4315223>] kernel_init_freeable+0x285/0x29a [ 27.316885] [<ffffffffb14472b0>] ? rest_init+0xd0/0xd0 [ 27.316885] [<ffffffffb14472be>] kernel_init+0xe/0xf0 [ 27.316885] [<ffffffffb148007c>] ret_from_fork+0x7c/0xb0 [ 27.316885] [<ffffffffb14472b0>] ? rest_init+0xd0/0xd0 [ 27.316885] Dumping ftrace buffer: [ 27.316885] (ftrace buffer empty) [ 27.316885] Kernel Offset: 0x2d000000 from 0xffffffff81000000 (relocation
Quoted 4 lines
range: 0xffffffff80000000-0xffffffffbfffffff) [ 27.316885] Rebooting in 1
seconds..virtio-9p is not even listed as an option now.
Thanks for reporting the problem, we're currently working on a fix:
* https://lkml.org/lkml/2015/1/20/710
Signature · 2 lines
paul moore security @ redhat
PATCH 3/5 audit: enable filename recording via getname_kernel()
In reply to Paul MooreEnable recording of filenames in getname_kernel() and remove the kludgy workaround in __audit_inode() now that we have proper filename logging for kernel users.
--- fs/namei.c | 1 + kernel/auditsc.c | 40 +++------------------------------------- 2 files changed, 4 insertions(+), 37 deletions(-)
fs/namei.c +1 -0
@@ -242,6 +242,7 @@ getname_kernel(const char * filename)strlcpy ((char *)result->name, filename, len); result->uptr =NULL ; result->aname =NULL ; +audit_getname (result);return result; }
kernel/auditsc.c +3 -37
@@ -1882,44 +1882,10 @@ out_alloc: n =audit_alloc_name (context,AUDIT_TYPE_UNKNOWN );if (!n)return ; -/* unfortunately, while we may have a path name to record with the -* inode, we can't always rely on the string lasting until the end of -* the syscall so we need to create our own copy, it may fail due to -* memory allocation issues, but we do our best */ -if (name) { -/* we can't use getname_kernel() due to size limits */ -size_t len =strlen (name->name) +1 ; -struct filename *new =__getname (); - -if (unlikely (!new)) - goto out; +if (name) +/* no need to set ->name_put as the original will cleanup */ + n->name = name; -if (len <= (PATH_MAX -sizeof (*new))) { - new->name = (char *)(new) +sizeof (*new); - new->separate = false; - }else if (len <=PATH_MAX ) { -/* this looks odd, but is due to final_putname() */ -struct filename *new2; - - new2 =kmalloc (sizeof (*new2),GFP_KERNEL ); -if (unlikely (!new2)) { -__putname (new); - goto out; - } - new2->name = (char *)new; - new2->separate = true; - new = new2; - }else { -/* we should never get here, but let's be safe */ -__putname (new); - goto out; - } -strlcpy ((char *)new->name, name->name, len); - new->uptr =NULL ; - new->aname = n; - n->name = new; - n->name_put = true; - } out:if (parent) { n->name_len = n->name ?parent_len (n->name->name) :AUDIT_NAME_FULL ;
PATCH 4/5 audit: fix filename matching in __audit_inode() and __audit_inode_child()
In reply to Paul MooreIn all likelihood there were some subtle, and perhaps not so subtle, bugs with filename matching in audit_inode() and audit_inode_child() for some time, however, recent changes to the audit filename code have definitely broken the filename matching code. The breakage could result in duplicate filenames in the audit log and other odd audit record entries. This patch fixes the filename matching code and restores some sanity to the filename audit records.
--- kernel/auditsc.c | 34 +++++++++++++++++++++++++--------- 1 file changed, 25 insertions(+), 9 deletions(-)
kernel/auditsc.c +25 -9
@@ -1846,6 +1846,7 @@ void __audit_inode(struct filename *name, const struct dentry *dentry,/* The struct filename _must_ have a populated ->name */ BUG_ON (!name->name);#endif +/* * If we have a pointer to an audit_names entry already, then we can * just use it directly if the type is correct.
@@ -1863,7 +1864,17 @@ void __audit_inode(struct filename *name, const struct dentry *dentry, }list_for_each_entry_reverse (n, &context->names_list, list) { -if (!n->name ||strcmp (n->name->name, name->name)) +if (inode->i_ino) { +/* valid inode number, use that for the comparison */ +if (n->ino != inode->i_ino || + n->dev != inode->i_sb->s_dev) +continue ; + }else if (n->name) { +/* inode number has not been set, check the name */ +if (strcmp (n->name->name, name->name)) +continue ; + }else +/* no inode and no name (?!) ... this is odd ... */ continue ;/* match the correct record type */
@@ -1931,11 +1942,16 @@ void __audit_inode_child(const struct inode *parent,/* look for a parent entry first */ list_for_each_entry (n, &context->names_list, list) { -if (!n->name || n->type !=AUDIT_TYPE_PARENT ) +if (!n->name || + (n->type !=AUDIT_TYPE_PARENT && + n->type !=AUDIT_TYPE_UNKNOWN ))continue ; -if (n->ino == parent->i_ino && - !audit_compare_dname_path (dname, n->name->name, n->name_len)) { +if (n->ino == parent->i_ino && n->dev == parent->i_sb->s_dev && + !audit_compare_dname_path (dname, + n->name->name, n->name_len)) { +if (n->type ==AUDIT_TYPE_UNKNOWN ) + n->type =AUDIT_TYPE_PARENT ; found_parent = n;break ; }
@@ -1944,11 +1960,8 @@ void __audit_inode_child(const struct inode *parent,/* is there a matching child entry? */ list_for_each_entry (n, &context->names_list, list) {/* can only match entries that have a name */ -if (!n->name || n->type != type) -continue ; - -/* if we found a parent, make sure this one is a child of it */ -if (found_parent && (n->name != found_parent->name)) +if (!n->name || + (n->type != type && n->type !=AUDIT_TYPE_UNKNOWN ))continue ; if (!strcmp(dname, n->name->name) ||
@@ -1956,6 +1969,8 @@ void __audit_inode_child(const struct inode *parent, found_parent ? found_parent->name_len : AUDIT_NAME_FULL)) { +if (n->type ==AUDIT_TYPE_UNKNOWN ) + n->type = type; found_child = n;break ; }
@@ -1984,6 +1999,7 @@ void __audit_inode_child(const struct inode *parent, found_child->name_put = false; } } +if (inode)audit_copy_inode (found_child, dentry, inode);else
PATCH 5/5 audit: replace getname()/putname() hacks with reference counters
In reply to Paul MooreIn order to ensure that filenames are not released before the audit subsystem is done with the strings there are a number of hacks built into the fs and audit subsystems around getname() and putname(). To say these hacks are "ugly" would be kind.
This patch removes the filename hackery in favor of a more conventional reference count based approach. The diffstat below tells most of the story; lots of audit/fs specific code is replaced with a traditional reference count based approach that is easily understood, even by those not familiar with the audit and/or fs subsystems.
--- fs/namei.c | 29 +++++++------- include/linux/audit.h | 3 - include/linux/fs.h | 9 +--- kernel/audit.h | 17 +------- kernel/auditsc.c | 101 ++++--------------------------------------------- 5 files changed, 26 insertions(+), 133 deletions(-)
fs/namei.c +14 -15
@@ -117,15 +117,6 @@* POSIX.1 2.4: an empty pathname is invalid (ENOENT). * PATH_MAX includes the nul terminator --RR. */ -void final_putname (struct filename *name) -{ -if (name->separate) { -__putname (name->name); -kfree (name); - }else { -__putname (name); - } -}#define EMBEDDED_NAME_MAX (PATH_MAX - sizeof(struct filename))
@@ -144,6 +135,7 @@ getname_flags(const char __user *filename, int flags, int *empty) result =__getname ();if (unlikely (!result))return ERR_PTR (-ENOMEM ); + result->refcnt =1 ;/* * First, try to embed the struct filename inside the names_cache
@@ -178,6 +170,7 @@ recopy: } result->name = kname; result->separate = true; + result->refcnt =1 ; max =PATH_MAX ; goto recopy; }
@@ -201,7 +194,7 @@ recopy:return result; error: -final_putname (result); +putname (result);return err; }
@@ -242,19 +235,25 @@ getname_kernel(const char * filename)strlcpy ((char *)result->name, filename, len); result->uptr =NULL ; result->aname =NULL ; + result->refcnt =1 ;audit_getname (result);return result; } -#ifdef CONFIG_AUDITSYSCALL void putname (struct filename *name) { -if (unlikely (!audit_dummy_context ())) -return audit_putname (name); -final_putname (name); +BUG_ON (name->refcnt <=0 ); + +if (--name->refcnt >0 ) +return ; + +if (name->separate) { +__putname (name->name); +kfree (name); + }else +__putname (name); } -#endif static int check_acl (struct inode *inode,int mask) {
include/linux/audit.h +0 -3
@@ -128,7 +128,6 @@ extern void __audit_syscall_entry(int major, unsigned long a0, unsigned long a1,extern void __audit_syscall_exit (int ret_success,long ret_value);extern struct filename *__audit_reusename (const __user char *uptr);extern void __audit_getname (struct filename *name); -extern void audit_putname (struct filename *name);#define AUDIT_INODE_PARENT 1/* dentry represents the parent */ #define AUDIT_INODE_HIDDEN 2/* audit record should be hidden */
@@ -347,8 +346,6 @@ static inline struct filename *audit_reusename(const __user char *name) }static inline void audit_getname (struct filename *name) { } -static inline void audit_putname (struct filename *name) -{ }static inline void __audit_inode (struct filename *name,const struct dentry *dentry,unsigned int flags)
include/linux/fs.h +2 -7
@@ -2017,6 +2017,7 @@ struct filename {const char *name;/* pointer to actual string */ const __user char *uptr;/* original userland pointer */ struct audit_names *aname; +int refcnt;bool separate;/* should "name" be freed? */ };
@@ -2036,6 +2037,7 @@ extern int filp_close(struct file *, fl_owner_t id);extern struct filename *getname (const char __user *);extern struct filename *getname_kernel (const char *); +extern void putname (struct filename *name);enum {FILE_CREATED =1 ,
@@ -2056,15 +2058,8 @@ extern void __init vfs_caches_init(unsigned long);extern struct kmem_cache *names_cachep; -extern void final_putname (struct filename *name); -#define __getname () kmem_cache_alloc(names_cachep, GFP_KERNEL)#define __putname (name) kmem_cache_free(names_cachep, (void *)(name)) -#ifndef CONFIG_AUDITSYSCALL -#define putname (name) final_putname(name) -#else -extern void putname (struct filename *name); -#endif #ifdef CONFIG_BLOCK extern int register_blkdev (unsigned int ,const char *);
kernel/audit.h +2 -15
@@ -24,12 +24,6 @@#include <linux/skbuff.h> #include <uapi/linux/mqueue.h> -/* 0 = no checking -1 = put_count checking -2 = verbose put_count checking -*/ -#define AUDIT_DEBUG 0 -/* AUDIT_NAMES is the number of slots we reserve in the audit_context * for saving names from getname(). If we get more names we will allocate * a name dynamically and also add those to the list anchored by names_list. */
@@ -74,9 +68,8 @@ struct audit_cap_data { }; }; -/* When fs/namei.c:getname() is called, we store the pointer in name and -* we don't let putname() free it (instead we free all of the saved -* pointers at syscall exit time). +/* When fs/namei.c:getname() is called, we store the pointer in name and bump +* the refcnt in the associated filename struct. * * Further, in fs/namei.c:path_lookup() we store the inode and device. */
@@ -86,7 +79,6 @@ struct audit_names {struct filename *name;int name_len;/* number of chars to log */ bool hidden;/* don't log this record */ -bool name_put;/* call __putname()? */ unsigned long ino;dev_t dev;
@@ -208,11 +200,6 @@ struct audit_context { };int fds[2 ];struct audit_proctitle proctitle; - -#if AUDIT_DEBUG -int put_count; -int ino_count; -#endif };extern u32 audit_ever_enabled;
kernel/auditsc.c +8 -93
153 lines not on this page · raw message