* Re: KASAN: use-after-free Read in tomoyo_realpath_from_path
[not found] <0000000000004f43fa058a97f4d3@google.com>
@ 2019-06-06 2:08 ` Tetsuo Handa
2019-06-06 5:20 ` Tetsuo Handa
1 sibling, 0 replies; 11+ messages in thread
From: Tetsuo Handa @ 2019-06-06 2:08 UTC (permalink / raw)
To: Al Viro, linux-fsdevel
Cc: syzbot, jmorris, linux-kernel, linux-security-module, serge,
syzkaller-bugs, takedakn
Here is a reproducer.
The problem is that TOMOYO is accessing already freed socket from security_file_open()
which later fails with -ENXIO (because we can't get file descriptor of sockets via
/proc/pid/fd/n interface), and the file descriptor is getting released before
security_file_open() completes because we do not raise "struct file"->f_count of
the file which is accessible via /proc/pid/fd/n interface. We can avoid this problem
if we can avoid calling security_file_open() which after all fails with -ENXIO.
How should we handle this race? Let LSM modules check if security_file_open() was
called on a socket?
----------------------------------------
diff --git a/fs/open.c b/fs/open.c
index b5b80469b93d..995ffcb37128 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -765,6 +765,12 @@ static int do_dentry_open(struct file *f,
error = security_file_open(f);
if (error)
goto cleanup_all;
+ if (!strcmp(current->comm, "a.out") &&
+ f->f_path.dentry->d_sb->s_magic == SOCKFS_MAGIC) {
+ printk("Start open(socket) delay\n");
+ schedule_timeout_killable(HZ * 5);
+ printk("End open(socket) delay\n");
+ }
error = break_lease(locks_inode(f), f->f_flags);
if (error)
----------------------------------------
----------------------------------------
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/socket.h>
int main(int argc, char *argv[])
{
pid_t pid = getpid();
int fd = socket(AF_ISDN, SOCK_RAW, 0);
char buffer[128] = { };
if (fork() == 0) {
close(fd);
snprintf(buffer, sizeof(buffer) - 1, "/proc/%u/fd/%u", pid, fd);
open(buffer, 3);
_exit(0);
}
sleep(2);
close(fd);
return 0;
}
----------------------------------------
----------------------------------------
getpid() = 32504
socket(AF_ISDN, SOCK_RAW, 0) = 3
clone(strace: Process 32505 attached
child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7efea30dda10) = 32505
[pid 32504] rt_sigprocmask(SIG_BLOCK, [CHLD], <unfinished ...>
[pid 32505] close(3 <unfinished ...>
[pid 32504] <... rt_sigprocmask resumed> [], 8) = 0
[pid 32505] <... close resumed> ) = 0
[pid 32504] rt_sigaction(SIGCHLD, NULL, {SIG_DFL, [], 0}, 8) = 0
[pid 32505] open("/proc/32504/fd/3", O_ACCMODE <unfinished ...>
[pid 32504] rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
[pid 32504] nanosleep({2, 0}, 0x7ffd3c608150) = 0
[pid 32504] close(3) = 0
[pid 32504] exit_group(0) = ?
[pid 32504] +++ exited with 0 +++
<... open resumed> ) = -1 ENXIO (No such device or address)
exit_group(0) = ?
----------------------------------------
----------------------------------------
[ 95.109628] Start open(socket) delay
[ 97.113150] base_sock_release(00000000506a3239) sk=00000000016d0ceb
[ 100.142235] End open(socket) delay
----------------------------------------
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: KASAN: use-after-free Read in tomoyo_realpath_from_path
[not found] <0000000000004f43fa058a97f4d3@google.com>
2019-06-06 2:08 ` KASAN: use-after-free Read in tomoyo_realpath_from_path Tetsuo Handa
@ 2019-06-06 5:20 ` Tetsuo Handa
2019-06-09 6:41 ` [PATCH] tomoyo: Don't check open/getattr permission on sockets Tetsuo Handa
1 sibling, 1 reply; 11+ messages in thread
From: Tetsuo Handa @ 2019-06-06 5:20 UTC (permalink / raw)
To: Al Viro, linux-fsdevel
Cc: syzbot, jmorris, linux-kernel, linux-security-module, serge,
syzkaller-bugs, takedakn
Tetsuo Handa wrote:
> The problem is that TOMOYO is accessing already freed socket from security_file_open()
> which later fails with -ENXIO (because we can't get file descriptor of sockets via
> /proc/pid/fd/n interface), and the file descriptor is getting released before
> security_file_open() completes because we do not raise "struct file"->f_count of
> the file which is accessible via /proc/pid/fd/n interface. We can avoid this problem
> if we can avoid calling security_file_open() which after all fails with -ENXIO.
> How should we handle this race? Let LSM modules check if security_file_open() was
> called on a socket?
Well, just refusing security_file_open() is not sufficient, for open(O_PATH) allows installing
file descriptor where SOCKET_I(inode)->sk can change at any moment, and TOMOYO cannot tell
whether it is safe to access SOCKET_I(inode)->sk from security_inode_getattr().
But refusing open(O_PATH) as well might break userspace programs. Oh, no...
----------------------------------------
diff --git a/fs/open.c b/fs/open.c
index b5b80469b93d..ea69668e2cd8 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -728,6 +728,16 @@ static int do_dentry_open(struct file *f,
/* Ensure that we skip any errors that predate opening of the file */
f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
+ /*
+ * Sockets must not be opened via /proc/pid/fd/n, even with O_PATH,
+ * for SOCKET_I(inode)->sk can be kfree()d at any moment after a file
+ * descriptor obtained by opening /proc/pid/fd/n was installed.
+ */
+ if (unlikely(S_ISSOCK(inode->i_mode))) {
+ error = (f->f_flags & O_PATH) ? -ENOENT : -ENXIO;
+ goto cleanup_file;
+ }
+
if (unlikely(f->f_flags & O_PATH)) {
f->f_mode = FMODE_PATH | FMODE_OPENED;
f->f_op = &empty_fops;
----------------------------------------
----------------------------------------
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/socket.h>
int main(int argc, char *argv[])
{
pid_t pid = getpid();
int fd = socket(AF_INET, SOCK_STREAM, 0);
char buffer[128] = { };
if (fork() == 0) {
struct stat buf = { };
close(fd);
snprintf(buffer, sizeof(buffer) - 1, "/proc/%u/fd/%u", pid, fd);
fd = open(buffer, __O_PATH);
sleep(5);
fstat(fd, &buf);
_exit(0);
}
sleep(2);
close(fd);
return 0;
}
----------------------------------------
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH] tomoyo: Don't check open/getattr permission on sockets.
2019-06-06 5:20 ` Tetsuo Handa
@ 2019-06-09 6:41 ` Tetsuo Handa
2019-06-16 6:49 ` Tetsuo Handa
0 siblings, 1 reply; 11+ messages in thread
From: Tetsuo Handa @ 2019-06-09 6:41 UTC (permalink / raw)
To: Al Viro, linux-fsdevel
Cc: syzbot, jmorris, linux-kernel, linux-security-module, serge,
syzkaller-bugs, takedakn
syzbot is reporting that use of SOCKET_I()->sk from open() can result in
use after free problem [1], for socket's inode is still reachable via
/proc/pid/fd/n despite destruction of SOCKET_I()->sk already completed.
But there is no point with calling security_file_open() on sockets
because open("/proc/pid/fd/n", !O_PATH) on sockets fails with -ENXIO.
There is some point with calling security_inode_getattr() on sockets
because stat("/proc/pid/fd/n") and fstat(open("/proc/pid/fd/n", O_PATH))
are valid. If we want to access "struct sock"->sk_{family,type,protocol}
fields, we will need to use security_socket_post_create() hook and
security_inode_free() hook in order to remember these fields because
security_sk_free() hook is called before the inode is destructed. But
since information which can be protected by checking
security_inode_getattr() on sockets is trivial, let's not be bothered by
"struct inode"->i_security management.
There is point with calling security_file_ioctl() on sockets. Since
ioctl(open("/proc/pid/fd/n", O_PATH)) is invalid, security_file_ioctl()
on sockets should remain safe.
[1] https://syzkaller.appspot.com/bug?id=73d590010454403d55164cca23bd0565b1eb3b74
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: syzbot <syzbot+0341f6a4d729d4e0acf1@syzkaller.appspotmail.com>
---
security/tomoyo/tomoyo.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/security/tomoyo/tomoyo.c b/security/tomoyo/tomoyo.c
index 716c92e..9661b86 100644
--- a/security/tomoyo/tomoyo.c
+++ b/security/tomoyo/tomoyo.c
@@ -126,6 +126,9 @@ static int tomoyo_bprm_check_security(struct linux_binprm *bprm)
*/
static int tomoyo_inode_getattr(const struct path *path)
{
+ /* It is not safe to call tomoyo_get_socket_name(). */
+ if (path->dentry->d_inode && S_ISSOCK(path->dentry->d_inode->i_mode))
+ return 0;
return tomoyo_path_perm(TOMOYO_TYPE_GETATTR, path, NULL);
}
@@ -316,6 +319,10 @@ static int tomoyo_file_open(struct file *f)
/* Don't check read permission here if called from do_execve(). */
if (current->in_execve)
return 0;
+ /* Sockets can't be opened by open(). */
+ if (f->f_path.dentry->d_inode &&
+ S_ISSOCK(f->f_path.dentry->d_inode->i_mode))
+ return 0;
return tomoyo_check_open_permission(tomoyo_domain(), &f->f_path,
f->f_flags);
}
--
1.8.3.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH] tomoyo: Don't check open/getattr permission on sockets.
2019-06-09 6:41 ` [PATCH] tomoyo: Don't check open/getattr permission on sockets Tetsuo Handa
@ 2019-06-16 6:49 ` Tetsuo Handa
2019-06-18 20:49 ` Al Viro
0 siblings, 1 reply; 11+ messages in thread
From: Tetsuo Handa @ 2019-06-16 6:49 UTC (permalink / raw)
To: Al Viro, linux-fsdevel
Cc: syzbot, jmorris, linux-kernel, linux-security-module, serge,
syzkaller-bugs, takedakn, David S. Miller
Hello, Al.
Q1: Do you agree that we should fix TOMOYO side rather than SOCKET_I()->sk
management.
Q2: Do you see any problem with using f->f_path.dentry->d_inode ?
Do we need to use d_backing_inode() or d_inode() ?
Regards.
On 2019/06/09 15:41, Tetsuo Handa wrote:
> syzbot is reporting that use of SOCKET_I()->sk from open() can result in
> use after free problem [1], for socket's inode is still reachable via
> /proc/pid/fd/n despite destruction of SOCKET_I()->sk already completed.
>
> But there is no point with calling security_file_open() on sockets
> because open("/proc/pid/fd/n", !O_PATH) on sockets fails with -ENXIO.
>
> There is some point with calling security_inode_getattr() on sockets
> because stat("/proc/pid/fd/n") and fstat(open("/proc/pid/fd/n", O_PATH))
> are valid. If we want to access "struct sock"->sk_{family,type,protocol}
> fields, we will need to use security_socket_post_create() hook and
> security_inode_free() hook in order to remember these fields because
> security_sk_free() hook is called before the inode is destructed. But
> since information which can be protected by checking
> security_inode_getattr() on sockets is trivial, let's not be bothered by
> "struct inode"->i_security management.
>
> There is point with calling security_file_ioctl() on sockets. Since
> ioctl(open("/proc/pid/fd/n", O_PATH)) is invalid, security_file_ioctl()
> on sockets should remain safe.
>
> [1] https://syzkaller.appspot.com/bug?id=73d590010454403d55164cca23bd0565b1eb3b74
>
> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> Reported-by: syzbot <syzbot+0341f6a4d729d4e0acf1@syzkaller.appspotmail.com>
> ---
> security/tomoyo/tomoyo.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/security/tomoyo/tomoyo.c b/security/tomoyo/tomoyo.c
> index 716c92e..9661b86 100644
> --- a/security/tomoyo/tomoyo.c
> +++ b/security/tomoyo/tomoyo.c
> @@ -126,6 +126,9 @@ static int tomoyo_bprm_check_security(struct linux_binprm *bprm)
> */
> static int tomoyo_inode_getattr(const struct path *path)
> {
> + /* It is not safe to call tomoyo_get_socket_name(). */
> + if (path->dentry->d_inode && S_ISSOCK(path->dentry->d_inode->i_mode))
> + return 0;
> return tomoyo_path_perm(TOMOYO_TYPE_GETATTR, path, NULL);
> }
>
> @@ -316,6 +319,10 @@ static int tomoyo_file_open(struct file *f)
> /* Don't check read permission here if called from do_execve(). */
> if (current->in_execve)
> return 0;
> + /* Sockets can't be opened by open(). */
> + if (f->f_path.dentry->d_inode &&
> + S_ISSOCK(f->f_path.dentry->d_inode->i_mode))
> + return 0;
> return tomoyo_check_open_permission(tomoyo_domain(), &f->f_path,
> f->f_flags);
> }
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] tomoyo: Don't check open/getattr permission on sockets.
2019-06-16 6:49 ` Tetsuo Handa
@ 2019-06-18 20:49 ` Al Viro
2019-06-22 4:45 ` [PATCH v2] " Tetsuo Handa
0 siblings, 1 reply; 11+ messages in thread
From: Al Viro @ 2019-06-18 20:49 UTC (permalink / raw)
To: Tetsuo Handa
Cc: linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
On Sun, Jun 16, 2019 at 03:49:00PM +0900, Tetsuo Handa wrote:
> Hello, Al.
>
> Q1: Do you agree that we should fix TOMOYO side rather than SOCKET_I()->sk
> management.
You do realize that sockets are not unique in that respect, right?
All kinds of interesting stuff can be accessed via /proc/*/fd/*, and
it _can_ be closed under you. So I'd suggest checking how your code
copes with similar for pipes, FIFOs, epoll, etc., accessed that way...
We are _not_ going to be checking that in fs/open.c - the stuff found
via /proc/*/fd/* can have the associated file closed by the time
we get to calling ->open() and we won't know that until said call.
> Q2: Do you see any problem with using f->f_path.dentry->d_inode ?
> Do we need to use d_backing_inode() or d_inode() ?
Huh? What's wrong with file_inode(f), in the first place? And
just when can that be NULL, while we are at it?
> > static int tomoyo_inode_getattr(const struct path *path)
> > {
> > + /* It is not safe to call tomoyo_get_socket_name(). */
> > + if (path->dentry->d_inode && S_ISSOCK(path->dentry->d_inode->i_mode))
> > + return 0;
Can that be called for a negative?
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
2019-06-18 20:49 ` Al Viro
@ 2019-06-22 4:45 ` Tetsuo Handa
2019-08-22 6:30 ` Eric Biggers
0 siblings, 1 reply; 11+ messages in thread
From: Tetsuo Handa @ 2019-06-22 4:45 UTC (permalink / raw)
To: Al Viro
Cc: linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
On 2019/06/19 5:49, Al Viro wrote:
> On Sun, Jun 16, 2019 at 03:49:00PM +0900, Tetsuo Handa wrote:
>> Hello, Al.
>>
>> Q1: Do you agree that we should fix TOMOYO side rather than SOCKET_I()->sk
>> management.
>
> You do realize that sockets are not unique in that respect, right?
> All kinds of interesting stuff can be accessed via /proc/*/fd/*, and
> it _can_ be closed under you. So I'd suggest checking how your code
> copes with similar for pipes, FIFOs, epoll, etc., accessed that way...
I know all kinds of interesting stuff can be accessed via /proc/*/fd/*,
and it _can_ be closed under me.
Regarding sockets, I was accessing "struct socket" memory and
"struct sock" memory which are outside of "struct inode" memory.
But regarding other objects, I am accessing "struct dentry" memory,
"struct super_block" memory and "struct inode" memory. I'm expecting
that these memory can't be kfree()d as long as "struct path" holds
a reference. Is my expectation correct?
>
> We are _not_ going to be checking that in fs/open.c - the stuff found
> via /proc/*/fd/* can have the associated file closed by the time
> we get to calling ->open() and we won't know that until said call.
OK. Then, fixing TOMOYO side is the correct way.
>
>> Q2: Do you see any problem with using f->f_path.dentry->d_inode ?
>> Do we need to use d_backing_inode() or d_inode() ?
>
> Huh? What's wrong with file_inode(f), in the first place? And
> just when can that be NULL, while we are at it?
Oh, I was not aware of file_inode(). Thanks.
>
>>> static int tomoyo_inode_getattr(const struct path *path)
>>> {
>>> + /* It is not safe to call tomoyo_get_socket_name(). */
>>> + if (path->dentry->d_inode && S_ISSOCK(path->dentry->d_inode->i_mode))
>>> + return 0;
>
> Can that be called for a negative?
>
I check for NULL when I'm not sure it is guaranteed to hold a valid pointer.
You meant "we are sure that path->dentry->d_inode is valid", don't you?
By the way, "negative" associates with IS_ERR() range. I guess that
"NULL" is the better name...
Anyway, here is V2 patch.
From c63c4074300921d6d1c33c3b8dc9c84ebfededf5 Mon Sep 17 00:00:00 2001
From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Date: Sat, 22 Jun 2019 13:14:26 +0900
Subject: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
syzbot is reporting that use of SOCKET_I()->sk from open() can result in
use after free problem [1], for socket's inode is still reachable via
/proc/pid/fd/n despite destruction of SOCKET_I()->sk already completed.
But there is no point with calling security_file_open() on sockets
because open("/proc/pid/fd/n", !O_PATH) on sockets fails with -ENXIO.
There is some point with calling security_inode_getattr() on sockets
because stat("/proc/pid/fd/n") and fstat(open("/proc/pid/fd/n", O_PATH))
are valid. If we want to access "struct sock"->sk_{family,type,protocol}
fields, we will need to use security_socket_post_create() hook and
security_inode_free() hook in order to remember these fields because
security_sk_free() hook is called before the inode is destructed. But
since information which can be protected by checking
security_inode_getattr() on sockets is trivial, let's not be bothered by
"struct inode"->i_security management.
There is point with calling security_file_ioctl() on sockets. Since
ioctl(open("/proc/pid/fd/n", O_PATH)) is invalid, security_file_ioctl()
on sockets should remain safe.
[1] https://syzkaller.appspot.com/bug?id=73d590010454403d55164cca23bd0565b1eb3b74
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: syzbot <syzbot+0341f6a4d729d4e0acf1@syzkaller.appspotmail.com>
---
security/tomoyo/tomoyo.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/security/tomoyo/tomoyo.c b/security/tomoyo/tomoyo.c
index 716c92e..8ea3f5d 100644
--- a/security/tomoyo/tomoyo.c
+++ b/security/tomoyo/tomoyo.c
@@ -126,6 +126,9 @@ static int tomoyo_bprm_check_security(struct linux_binprm *bprm)
*/
static int tomoyo_inode_getattr(const struct path *path)
{
+ /* It is not safe to call tomoyo_get_socket_name(). */
+ if (S_ISSOCK(d_inode(path->dentry)->i_mode))
+ return 0;
return tomoyo_path_perm(TOMOYO_TYPE_GETATTR, path, NULL);
}
@@ -316,6 +319,9 @@ static int tomoyo_file_open(struct file *f)
/* Don't check read permission here if called from do_execve(). */
if (current->in_execve)
return 0;
+ /* Sockets can't be opened by open(). */
+ if (S_ISSOCK(file_inode(f)->i_mode))
+ return 0;
return tomoyo_check_open_permission(tomoyo_domain(), &f->f_path,
f->f_flags);
}
--
1.8.3.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
2019-06-22 4:45 ` [PATCH v2] " Tetsuo Handa
@ 2019-08-22 6:30 ` Eric Biggers
2019-08-22 6:55 ` Tetsuo Handa
0 siblings, 1 reply; 11+ messages in thread
From: Eric Biggers @ 2019-08-22 6:30 UTC (permalink / raw)
To: Tetsuo Handa
Cc: Al Viro, linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
Hi Tetsuo,
On Sat, Jun 22, 2019 at 01:45:30PM +0900, Tetsuo Handa wrote:
> On 2019/06/19 5:49, Al Viro wrote:
> > On Sun, Jun 16, 2019 at 03:49:00PM +0900, Tetsuo Handa wrote:
> >> Hello, Al.
> >>
> >> Q1: Do you agree that we should fix TOMOYO side rather than SOCKET_I()->sk
> >> management.
> >
> > You do realize that sockets are not unique in that respect, right?
> > All kinds of interesting stuff can be accessed via /proc/*/fd/*, and
> > it _can_ be closed under you. So I'd suggest checking how your code
> > copes with similar for pipes, FIFOs, epoll, etc., accessed that way...
>
> I know all kinds of interesting stuff can be accessed via /proc/*/fd/*,
> and it _can_ be closed under me.
>
> Regarding sockets, I was accessing "struct socket" memory and
> "struct sock" memory which are outside of "struct inode" memory.
>
> But regarding other objects, I am accessing "struct dentry" memory,
> "struct super_block" memory and "struct inode" memory. I'm expecting
> that these memory can't be kfree()d as long as "struct path" holds
> a reference. Is my expectation correct?
>
> >
> > We are _not_ going to be checking that in fs/open.c - the stuff found
> > via /proc/*/fd/* can have the associated file closed by the time
> > we get to calling ->open() and we won't know that until said call.
>
> OK. Then, fixing TOMOYO side is the correct way.
>
> >
> >> Q2: Do you see any problem with using f->f_path.dentry->d_inode ?
> >> Do we need to use d_backing_inode() or d_inode() ?
> >
> > Huh? What's wrong with file_inode(f), in the first place? And
> > just when can that be NULL, while we are at it?
>
> Oh, I was not aware of file_inode(). Thanks.
>
> >
> >>> static int tomoyo_inode_getattr(const struct path *path)
> >>> {
> >>> + /* It is not safe to call tomoyo_get_socket_name(). */
> >>> + if (path->dentry->d_inode && S_ISSOCK(path->dentry->d_inode->i_mode))
> >>> + return 0;
> >
> > Can that be called for a negative?
> >
>
> I check for NULL when I'm not sure it is guaranteed to hold a valid pointer.
> You meant "we are sure that path->dentry->d_inode is valid", don't you?
>
> By the way, "negative" associates with IS_ERR() range. I guess that
> "NULL" is the better name...
>
> Anyway, here is V2 patch.
>
> From c63c4074300921d6d1c33c3b8dc9c84ebfededf5 Mon Sep 17 00:00:00 2001
> From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> Date: Sat, 22 Jun 2019 13:14:26 +0900
> Subject: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
>
> syzbot is reporting that use of SOCKET_I()->sk from open() can result in
> use after free problem [1], for socket's inode is still reachable via
> /proc/pid/fd/n despite destruction of SOCKET_I()->sk already completed.
>
> But there is no point with calling security_file_open() on sockets
> because open("/proc/pid/fd/n", !O_PATH) on sockets fails with -ENXIO.
>
> There is some point with calling security_inode_getattr() on sockets
> because stat("/proc/pid/fd/n") and fstat(open("/proc/pid/fd/n", O_PATH))
> are valid. If we want to access "struct sock"->sk_{family,type,protocol}
> fields, we will need to use security_socket_post_create() hook and
> security_inode_free() hook in order to remember these fields because
> security_sk_free() hook is called before the inode is destructed. But
> since information which can be protected by checking
> security_inode_getattr() on sockets is trivial, let's not be bothered by
> "struct inode"->i_security management.
>
> There is point with calling security_file_ioctl() on sockets. Since
> ioctl(open("/proc/pid/fd/n", O_PATH)) is invalid, security_file_ioctl()
> on sockets should remain safe.
>
> [1] https://syzkaller.appspot.com/bug?id=73d590010454403d55164cca23bd0565b1eb3b74
>
> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> Reported-by: syzbot <syzbot+0341f6a4d729d4e0acf1@syzkaller.appspotmail.com>
> ---
> security/tomoyo/tomoyo.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/security/tomoyo/tomoyo.c b/security/tomoyo/tomoyo.c
> index 716c92e..8ea3f5d 100644
> --- a/security/tomoyo/tomoyo.c
> +++ b/security/tomoyo/tomoyo.c
> @@ -126,6 +126,9 @@ static int tomoyo_bprm_check_security(struct linux_binprm *bprm)
> */
> static int tomoyo_inode_getattr(const struct path *path)
> {
> + /* It is not safe to call tomoyo_get_socket_name(). */
> + if (S_ISSOCK(d_inode(path->dentry)->i_mode))
> + return 0;
> return tomoyo_path_perm(TOMOYO_TYPE_GETATTR, path, NULL);
> }
>
> @@ -316,6 +319,9 @@ static int tomoyo_file_open(struct file *f)
> /* Don't check read permission here if called from do_execve(). */
> if (current->in_execve)
> return 0;
> + /* Sockets can't be opened by open(). */
> + if (S_ISSOCK(file_inode(f)->i_mode))
> + return 0;
> return tomoyo_check_open_permission(tomoyo_domain(), &f->f_path,
> f->f_flags);
> }
> --
What happened to this patch?
Also, isn't the same bug in other places too?:
- tomoyo_path_chmod()
- tomoyo_path_chown()
- smack_inode_getsecurity()
- smack_inode_setsecurity()
- Eric
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
2019-08-22 6:30 ` Eric Biggers
@ 2019-08-22 6:55 ` Tetsuo Handa
2019-08-22 7:01 ` Eric Biggers
0 siblings, 1 reply; 11+ messages in thread
From: Tetsuo Handa @ 2019-08-22 6:55 UTC (permalink / raw)
To: Eric Biggers
Cc: Al Viro, linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
Eric Biggers wrote:
> What happened to this patch?
I have to learn how to manage a git tree for sending
pull requests, but I can't find time to try.
>
> Also, isn't the same bug in other places too?:
>
> - tomoyo_path_chmod()
> - tomoyo_path_chown()
> - smack_inode_getsecurity()
> - smack_inode_setsecurity()
What's the bug? The file descriptor returned by open(O_PATH) cannot be
passed to read(2), write(2), fchmod(2), fchown(2), fgetxattr(2), mmap(2) etc.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
2019-08-22 6:55 ` Tetsuo Handa
@ 2019-08-22 7:01 ` Eric Biggers
2019-08-22 7:42 ` Tetsuo Handa
0 siblings, 1 reply; 11+ messages in thread
From: Eric Biggers @ 2019-08-22 7:01 UTC (permalink / raw)
To: Tetsuo Handa
Cc: Al Viro, linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
On Thu, Aug 22, 2019 at 03:55:31PM +0900, Tetsuo Handa wrote:
> Eric Biggers wrote:
> > What happened to this patch?
>
> I have to learn how to manage a git tree for sending
> pull requests, but I can't find time to try.
>
> >
> > Also, isn't the same bug in other places too?:
> >
> > - tomoyo_path_chmod()
> > - tomoyo_path_chown()
> > - smack_inode_getsecurity()
> > - smack_inode_setsecurity()
>
> What's the bug? The file descriptor returned by open(O_PATH) cannot be
> passed to read(2), write(2), fchmod(2), fchown(2), fgetxattr(2), mmap(2) etc.
>
chmod(2), chown(2), getxattr(2), and setxattr(2) take a path, not a fd.
- Eric
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
2019-08-22 7:01 ` Eric Biggers
@ 2019-08-22 7:42 ` Tetsuo Handa
2019-08-22 15:47 ` Eric Biggers
0 siblings, 1 reply; 11+ messages in thread
From: Tetsuo Handa @ 2019-08-22 7:42 UTC (permalink / raw)
To: Eric Biggers
Cc: Al Viro, linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
Eric Biggers wrote:
> On Thu, Aug 22, 2019 at 03:55:31PM +0900, Tetsuo Handa wrote:
> > > Also, isn't the same bug in other places too?:
> > >
> > > - tomoyo_path_chmod()
> > > - tomoyo_path_chown()
> > > - smack_inode_getsecurity()
> > > - smack_inode_setsecurity()
> >
> > What's the bug? The file descriptor returned by open(O_PATH) cannot be
> > passed to read(2), write(2), fchmod(2), fchown(2), fgetxattr(2), mmap(2) etc.
> >
>
> chmod(2), chown(2), getxattr(2), and setxattr(2) take a path, not a fd.
>
OK. Then, is the correct fix
inode_lock(inode);
if (SOCKET_I(inode)->sk) {
// Can access SOCKET_I(sock)->sk->*
} else {
// Already close()d. Don't touch.
}
inode_unlock(inode);
thanks to
commit 6d8c50dcb029872b ("socket: close race condition between sock_close() and sockfs_setattr()")
commit ff7b11aa481f682e ("net: socket: set sock->sk to NULL after calling proto_ops::release()")
changes?
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2] tomoyo: Don't check open/getattr permission on sockets.
2019-08-22 7:42 ` Tetsuo Handa
@ 2019-08-22 15:47 ` Eric Biggers
0 siblings, 0 replies; 11+ messages in thread
From: Eric Biggers @ 2019-08-22 15:47 UTC (permalink / raw)
To: Tetsuo Handa
Cc: Al Viro, linux-fsdevel, syzbot, jmorris, linux-kernel,
linux-security-module, serge, syzkaller-bugs, takedakn,
David S. Miller
On Thu, Aug 22, 2019 at 04:42:26PM +0900, Tetsuo Handa wrote:
> Eric Biggers wrote:
> > On Thu, Aug 22, 2019 at 03:55:31PM +0900, Tetsuo Handa wrote:
> > > > Also, isn't the same bug in other places too?:
> > > >
> > > > - tomoyo_path_chmod()
> > > > - tomoyo_path_chown()
> > > > - smack_inode_getsecurity()
> > > > - smack_inode_setsecurity()
> > >
> > > What's the bug? The file descriptor returned by open(O_PATH) cannot be
> > > passed to read(2), write(2), fchmod(2), fchown(2), fgetxattr(2), mmap(2) etc.
> > >
> >
> > chmod(2), chown(2), getxattr(2), and setxattr(2) take a path, not a fd.
> >
>
> OK. Then, is the correct fix
>
> inode_lock(inode);
> if (SOCKET_I(inode)->sk) {
> // Can access SOCKET_I(sock)->sk->*
> } else {
> // Already close()d. Don't touch.
> }
> inode_unlock(inode);
>
> thanks to
>
> commit 6d8c50dcb029872b ("socket: close race condition between sock_close() and sockfs_setattr()")
> commit ff7b11aa481f682e ("net: socket: set sock->sk to NULL after calling proto_ops::release()")
>
> changes?
inode_lock() is already held during security_path_chmod(),
security_path_chown(), and security_inode_setxattr().
So you can't just take it again.
- Eric
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2019-08-22 15:48 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <0000000000004f43fa058a97f4d3@google.com>
2019-06-06 2:08 ` KASAN: use-after-free Read in tomoyo_realpath_from_path Tetsuo Handa
2019-06-06 5:20 ` Tetsuo Handa
2019-06-09 6:41 ` [PATCH] tomoyo: Don't check open/getattr permission on sockets Tetsuo Handa
2019-06-16 6:49 ` Tetsuo Handa
2019-06-18 20:49 ` Al Viro
2019-06-22 4:45 ` [PATCH v2] " Tetsuo Handa
2019-08-22 6:30 ` Eric Biggers
2019-08-22 6:55 ` Tetsuo Handa
2019-08-22 7:01 ` Eric Biggers
2019-08-22 7:42 ` Tetsuo Handa
2019-08-22 15:47 ` Eric Biggers
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).