file capabilities: remove cap_task_kill() (-git)
Subsystems (3)
PATCH 1/1 file capabilities: remove cap_task_kill() (-git)
(resending once against -git. I had sent against -stable in
http://lkml.org/lkml/2008/2/28/225. Without this patch, atd is broken at least on some distros.)
The original justification for cap_task_kill() was as follows:
check_kill_permission() does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.
However now that cap_task_kill() always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return -EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check in check_kill_permission().
One example of a still-broken application is 'at' for non-root users.
This patch removes cap_task_kill().
--- include/linux/security.h | 3 +-- security/capability.c | 1 - security/commoncap.c | 40 ---------------------------------------- 3 files changed, 1 insertions(+), 43 deletions(-)
include/linux/security.h +1 -2
@@ -57,7 +57,6 @@ extern int cap_inode_need_killpriv(struct dentry *dentry);extern int cap_inode_killpriv (struct dentry *dentry);extern int cap_task_post_setuid (uid_t old_ruid,uid_t old_euid,uid_t old_suid,int flags);extern void cap_task_reparent_to_init (struct task_struct *p); -extern int cap_task_kill (struct task_struct *p,struct siginfo *info,int sig,u32 secid);extern int cap_task_setscheduler (struct task_struct *p,int policy,struct sched_param *lp);extern int cap_task_setioprio (struct task_struct *p,int ioprio);extern int cap_task_setnice (struct task_struct *p,int nice);
@@ -2187,7 +2186,7 @@ static inline int security_task_kill (struct task_struct *p, struct siginfo *info, int sig, u32 secid) { - return cap_task_kill(p, info, sig, secid); + return 0; }static inline int security_task_wait (struct task_struct *p)
security/capability.c +0 -1
@@ -40,7 +40,6 @@ static struct security_operations capability_ops = { .inode_need_killpriv = cap_inode_need_killpriv, .inode_killpriv = cap_inode_killpriv, - .task_kill = cap_task_kill, .task_setscheduler = cap_task_setscheduler, .task_setioprio = cap_task_setioprio, .task_setnice = cap_task_setnice,
security/commoncap.c +0 -40
@@ -540,41 +540,6 @@ int cap_task_setnice (struct task_struct *p, int nice)return cap_safe_nice (p); } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -if (info !=SEND_SIG_NOINFO && (is_si_special (info) ||SI_FROMKERNEL (info))) -return 0 ; - -/* -* Running a setuid root program raises your capabilities. -* Killing your own setuid root processes was previously -* allowed. -* We must preserve legacy signal behavior in this case. -*/ -if (p->uid == current->uid) -return 0 ; - -/* sigcont is permitted within same session */ -if (sig ==SIGCONT && (task_session_nr (current) ==task_session_nr (p))) -return 0 ; - -if (secid) -/* -* Signal sent as a particular user. -* Capabilities are ignored. May be wrong, but it's the -* only thing we can do at the moment. -* Used only by usb drivers? -*/ -return 0 ; -if (cap_issubset (p->cap_permitted, current->cap_permitted)) -return 0 ; -if (capable (CAP_KILL )) -return 0 ; - -return -EPERM ; -} -/* * called from kernel/sys.c for prctl(PR_CABSET_DROP) * done without task_capability_lock() because it introduces
@@ -605,11 +570,6 @@ int cap_task_setnice (struct task_struct *p, int nice) {return 0 ; } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -return 0 ; -}#endif void cap_task_reparent_to_init (struct task_struct *p)
Signature · 1 line
1.5.2.5
Em Wed, 19 Mar 2008 11:56:35 -0500
"Serge E. Hallyn" <[email protected]> escreveu:
| (resending once against -git. I had sent against -stable in | http://lkml.org/lkml/2008/2/28/225. Without this patch, | atd is broken at least on some distros.) | | The original justification for cap_task_kill() was as follows: | | check_kill_permission() does appropriate uid equivalence checks. | However with file capabilities it becomes possible for an | unprivileged user to execute a file with file capabilities | resulting in a more privileged task with the same uid. | | However now that cap_task_kill() always returns 0 (permission | granted) when p->uid==current->uid, the whole hook is worthless, | and only likely to create more subtle problems in the corner cases | where it might still be called but return -EPERM. Those cases | are basically when uids are different but euid/suid is equivalent | as per the check in check_kill_permission(). | | One example of a still-broken application is 'at' for non-root users. | | This patch removes cap_task_kill(). | | Signed-off-by: Serge Hallyn <[email protected]> | Acked-by: Andrew G. Morgan <[email protected]>
I've tested the -stable patch (no time to test this one, but it's
exactly the same patch), not sure if this applies to this but...
Signature · 1 line
Luiz Fernando N. Capitulino
Quoted 3 lines · On Wed, 19 Mar 2008 11:56:35 -0500 "Serge E. Hallyn" <[email protected]> wrote:
(resending once against -git. I had sent against -stable in
http://lkml.org/lkml/2008/2/28/225. Without this patch, atd is broken at least on some distros.)
So this fix is needed in 2.6.24.x?
Quoted 7 lines
The original justification for
cap_task_kill()was as follows:
check_kill_permission()does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.
Quoted 8 lines
However now that
cap_task_kill()always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return-EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check incheck_kill_permission().One example of a still-broken application is 'at' for non-root users.
This 2.6.25-rc6 patch doesn't apply correctly to 2.6.24. I can't find a *formal* copy of your 2.6.24 patch on [email protected], so perhaps a resend for -stable is in order.
Quoted 8 lines · Andrew Morton
Quoted 3 lines · On Wed, 19 Mar 2008 11:56:35 -0500 "Serge E. Hallyn" <[email protected]> wrote:
(resending once against -git. I had sent against -stable in
http://lkml.org/lkml/2008/2/28/225. Without this patch, atd is broken at least on some distros.)So this fix is needed in 2.6.24.x?
Yes it is.
Quoted 14 lines
Quoted 14 lines
The original justification for
cap_task_kill()was as follows:
check_kill_permission()does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.However now that
cap_task_kill()always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return-EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check incheck_kill_permission().
Quoted 5 lines
Quoted 1 line
One example of a still-broken application is 'at' for non-root users.
This 2.6.25-rc6 patch doesn't apply correctly to 2.6.24. I can't find a *formal* copy of your 2.6.24 patch on [email protected], so perhaps a resend for -stable is in order.
Argh, yes, will do. Thanks.
-serge
Quoted 18 lines · On Wed, 19 Mar 2008 11:56:35 -0500 "Serge E. Hallyn" <[email protected]> wrote:
(resending once against -git. I had sent against -stable in
http://lkml.org/lkml/2008/2/28/225. Without this patch, atd is broken at least on some distros.)The original justification for
cap_task_kill()was as follows:
check_kill_permission()does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.However now that
cap_task_kill()always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return-EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check incheck_kill_permission().
Quoted 3 lines
One example of a still-broken application is 'at' for non-root users.
This patch removes
cap_task_kill().
umm,
security/smack/smack_lsm.c: In function 'smack_task_kill':security/smack/smack_lsm.c:1122: error: implicit declaration of function 'cap_task_kill'
Quoted 5 lines · Andrew Morton
umm,
security/smack/smack_lsm.c: In function 'smack_task_kill':security/smack/smack_lsm.c:1122: error: implicit declaration of function 'cap_task_kill'
Serge, can you resend with that fixed and the tested-by added?
Linus
Quoted 8 lines · Linus Torvalds
Quoted 5 lines · Andrew Morton
umm,
security/smack/smack_lsm.c: In function 'smack_task_kill':security/smack/smack_lsm.c:1122: error: implicit declaration of function 'cap_task_kill'
Right, that was againstgit://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
which doesn't yet have smack. I should've been clear about that.
Quoted 3 lines
Serge, can you resend with that fixed and the tested-by added?
Linus
Following is the version against this morning's mmotm with the tested-by added.
thanks,
-serge
Quoted 1 line
From
c50b1c9f7a9e9434c8ddb50cb81e6b342638b8e0Mon Sep 17 00:00:00 2001
From: Serge Hallyn <[email protected]>
Date: Fri, 29 Feb 2008 15:14:57 +0000
Subject: [PATCH 1/1] file capabilities: remove cap_task_kill() (-mmotm)
The original justification for cap_task_kill() was as follows:
check_kill_permission() does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.
However now that cap_task_kill() always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return -EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check in check_kill_permission().
One example of a still-broken application is 'at' for non-root users.
This patch removes cap_task_kill().
--- include/linux/security.h | 3 +-- security/capability.c | 1 - security/commoncap.c | 33 --------------------------------- security/smack/smack_lsm.c | 5 ----- 4 files changed, 1 insertions(+), 41 deletions(-)
include/linux/security.h +1 -2
@@ -59,7 +59,6 @@ extern int cap_inode_need_killpriv(struct dentry *dentry);extern int cap_inode_killpriv (struct dentry *dentry);extern int cap_task_post_setuid (uid_t old_ruid,uid_t old_euid,uid_t old_suid,int flags);extern void cap_task_reparent_to_init (struct task_struct *p); -extern int cap_task_kill (struct task_struct *p,struct siginfo *info,int sig,u32 secid);extern int cap_task_prctl (int option,unsigned long arg2,unsigned long arg3,unsigned long arg4,unsigned long arg5,long *rc_p);extern int cap_task_setscheduler (struct task_struct *p,int policy,struct sched_param *lp);
@@ -2276,7 +2275,7 @@ static inline int security_task_kill (struct task_struct *p, struct siginfo *info, int sig, u32 secid) { - return cap_task_kill(p, info, sig, secid); + return 0; }static inline int security_task_wait (struct task_struct *p)
security/capability.c +0 -1
@@ -40,7 +40,6 @@ static struct security_operations capability_ops = { .inode_need_killpriv = cap_inode_need_killpriv, .inode_killpriv = cap_inode_killpriv, - .task_kill = cap_task_kill, .task_setscheduler = cap_task_setscheduler, .task_setioprio = cap_task_setioprio, .task_setnice = cap_task_setnice,
security/commoncap.c +0 -33
@@ -537,34 +537,6 @@ int cap_task_setnice (struct task_struct *p, int nice)return cap_safe_nice (p); } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -/* -* Running a setuid root program raises your capabilities. -* Killing your own setuid root processes was previously -* allowed. -* We must preserve legacy signal behavior in this case. -*/ -if (p->uid == current->uid) -return 0 ; - -if (secid) -/* -* Signal sent as a particular user. -* Capabilities are ignored. May be wrong, but it's the -* only thing we can do at the moment. -* Used only by usb drivers? -*/ -return 0 ; -if (cap_issubset (p->cap_permitted, current->cap_permitted)) -return 0 ; -if (capable (CAP_KILL )) -return 0 ; - -return -EPERM ; -} -/* * called from kernel/sys.c for prctl(PR_CABSET_DROP) * done without task_capability_lock() because it introduces
@@ -596,11 +568,6 @@ int cap_task_setnice (struct task_struct *p, int nice) {return 0 ; } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -return 0 ; -}#endif int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
security/smack/smack_lsm.c +0 -5
@@ -1117,11 +1117,6 @@ static int smack_task_movememory(struct task_struct *p)static int smack_task_kill (struct task_struct *p,struct siginfo *info,int sig,u32 secid) { -int rc; - - rc =cap_task_kill (p, info, sig, secid); -if (rc !=0 ) -return rc;/* * Sending a signal requires that the sender * can write the receiver.
Signature · 1 line
1.5.2.5
Quoted 4 lines · Serge E. Hallyn
Right, that was against
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
which doesn't yet have smack. I should've been clear about that.
Ok, now I'm _really_ confused. I have smack in my tree, it got merged before -rc1.
So any patch that is against some version without smack is not a patch against a -git tree for the last several weeks.
Me confused.
Linus
Serge E. Hallyn
Acked-by Andrew G. MorganQuoted 10 lines · Linus Torvalds
Quoted 4 lines · Serge E. Hallyn
Right, that was against
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
which doesn't yet have smack. I should've been clear about that.Ok, now I'm _really_ confused. I have smack in my tree, it got merged before -rc1.
No clearly I'm the one confused/on drugs.
I see it now. I swear I checked my own freshly pulled tree and gitweb several times...
Quoted 6 lines
So any patch that is against some version without smack is not a patch against a -git tree for the last several weeks.
Me confused.
Linus
New patch against -git attached.
thanks,
-serge
Quoted 1 line
From
a0e56351e00b7b7442723b7ca6247c267c2628fdMon Sep 17 00:00:00 2001
From: Serge Hallyn <[email protected]>
Date: Fri, 29 Feb 2008 15:14:57 +0000
Subject: [PATCH 1/1] file capabilities: remove cap_task_kill() (-git)
The original justification for cap_task_kill() was as follows:
check_kill_permission() does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.
However now that cap_task_kill() always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return -EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check in check_kill_permission().
One example of a still-broken application is 'at' for non-root users.
This patch removes cap_task_kill().
--- include/linux/security.h | 3 +-- security/capability.c | 1 - security/commoncap.c | 40 ---------------------------------------- security/smack/smack_lsm.c | 5 ----- 4 files changed, 1 insertions(+), 48 deletions(-)
include/linux/security.h +1 -2
@@ -57,7 +57,6 @@ extern int cap_inode_need_killpriv(struct dentry *dentry);extern int cap_inode_killpriv (struct dentry *dentry);extern int cap_task_post_setuid (uid_t old_ruid,uid_t old_euid,uid_t old_suid,int flags);extern void cap_task_reparent_to_init (struct task_struct *p); -extern int cap_task_kill (struct task_struct *p,struct siginfo *info,int sig,u32 secid);extern int cap_task_setscheduler (struct task_struct *p,int policy,struct sched_param *lp);extern int cap_task_setioprio (struct task_struct *p,int ioprio);extern int cap_task_setnice (struct task_struct *p,int nice);
@@ -2187,7 +2186,7 @@ static inline int security_task_kill (struct task_struct *p, struct siginfo *info, int sig, u32 secid) { - return cap_task_kill(p, info, sig, secid); + return 0; }static inline int security_task_wait (struct task_struct *p)
security/capability.c +0 -1
@@ -40,7 +40,6 @@ static struct security_operations capability_ops = { .inode_need_killpriv = cap_inode_need_killpriv, .inode_killpriv = cap_inode_killpriv, - .task_kill = cap_task_kill, .task_setscheduler = cap_task_setscheduler, .task_setioprio = cap_task_setioprio, .task_setnice = cap_task_setnice,
security/commoncap.c +0 -40
@@ -540,41 +540,6 @@ int cap_task_setnice (struct task_struct *p, int nice)return cap_safe_nice (p); } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -if (info !=SEND_SIG_NOINFO && (is_si_special (info) ||SI_FROMKERNEL (info))) -return 0 ; - -/* -* Running a setuid root program raises your capabilities. -* Killing your own setuid root processes was previously -* allowed. -* We must preserve legacy signal behavior in this case. -*/ -if (p->uid == current->uid) -return 0 ; - -/* sigcont is permitted within same session */ -if (sig ==SIGCONT && (task_session_nr (current) ==task_session_nr (p))) -return 0 ; - -if (secid) -/* -* Signal sent as a particular user. -* Capabilities are ignored. May be wrong, but it's the -* only thing we can do at the moment. -* Used only by usb drivers? -*/ -return 0 ; -if (cap_issubset (p->cap_permitted, current->cap_permitted)) -return 0 ; -if (capable (CAP_KILL )) -return 0 ; - -return -EPERM ; -} -/* * called from kernel/sys.c for prctl(PR_CABSET_DROP) * done without task_capability_lock() because it introduces
@@ -605,11 +570,6 @@ int cap_task_setnice (struct task_struct *p, int nice) {return 0 ; } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -return 0 ; -}#endif void cap_task_reparent_to_init (struct task_struct *p)
security/smack/smack_lsm.c +0 -5
@@ -1117,11 +1117,6 @@ static int smack_task_movememory(struct task_struct *p)static int smack_task_kill (struct task_struct *p,struct siginfo *info,int sig,u32 secid) { -int rc; - - rc =cap_task_kill (p, info, sig, secid); -if (rc !=0 ) -return rc;/* * Special cases where signals really ought to go through * in spite of policy. Stephen Smalley suggests it may
Signature · 1 line
1.5.2.5
Quoted 45 lines · --- "Serge E. Hallyn
Quoted 8 lines · Linus Torvalds
Quoted 5 lines · Andrew Morton
umm,
security/smack/smack_lsm.c: In function 'smack_task_kill':security/smack/smack_lsm.c:1122: error: implicit declaration of function'
cap_task_kill'Right, that was against
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
which doesn't yet have smack. I should've been clear about that.Quoted 3 lines
Serge, can you resend with that fixed and the tested-by added?
Linus
Following is the version against this morning's mmotm with the tested-by added.
thanks,
-sergeQuoted 1 line
From
c50b1c9f7a9e9434c8ddb50cb81e6b342638b8e0Mon Sep 17 00:00:00 2001From: Serge Hallyn <[email protected]>
Date: Fri, 29 Feb 2008 15:14:57 +0000
Subject: [PATCH 1/1] file capabilities: removecap_task_kill()(-mmotm)The original justification for
cap_task_kill()was as follows:
check_kill_permission()does appropriate uid equivalence checks. However with file capabilities it becomes possible for an unprivileged user to execute a file with file capabilities resulting in a more privileged task with the same uid.However now that
cap_task_kill()always returns 0 (permission granted) when p->uid==current->uid, the whole hook is worthless, and only likely to create more subtle problems in the corner cases where it might still be called but return-EPERM. Those cases are basically when uids are different but euid/suid is equivalent as per the check incheck_kill_permission().
Quoted 7 lines
One example of a still-broken application is 'at' for non-root users.
This patch removes
cap_task_kill().Signed-off-by Serge Hallyn <[email protected]>Acked-by Andrew G. Morgan <[email protected]>
Quoted 112 lines
--- include/linux/security.h | 3 +-- security/capability.c | 1 - security/commoncap.c | 33 --------------------------------- security/smack/smack_lsm.c | 5 ----- 4 files changed, 1 insertions(+), 41 deletions(-)include/linux/security.h +0
-1@@ -59,7 +59,6 @@ extern int cap_inode_need_killpriv(struct dentry *dentry);extern int cap_inode_killpriv (struct dentry *dentry);extern int cap_task_post_setuid (uid_t old_ruid,uid_t old_euid,uid_t old_suid,int flags);extern void cap_task_reparent_to_init (struct task_struct *p); -extern int cap_task_kill (struct task_struct *p,struct siginfo *info,int sig, u32 secid); extern int cap_task_prctl(int option, unsigned long arg2, unsigned longarg3,
unsigned long arg4, unsigned long arg5, long*rc_p);
extern intcap_task_setscheduler(structtask_struct*p, int policy, structsched_param*lp);include/linux/security.h +1
-1@@ -2276,7 +2275,7 @@ static inline int security_task_kill (structtask_struct *p, struct siginfo *info, int sig, u32 secid) { - return cap_task_kill(p, info, sig, secid); + return 0; }static inline int
security_task_wait(structtask_struct*p)security/capability.c +0
-1@@ -40,7 +40,6 @@ static struct security_operations capability_ops = { .inode_need_killpriv = cap_inode_need_killpriv, .inode_killpriv = cap_inode_killpriv, - .task_kill = cap_task_kill, .task_setscheduler = cap_task_setscheduler, .task_setioprio = cap_task_setioprio, .task_setnice = cap_task_setnice,security/commoncap.c +0
-33@@ -537,34 +537,6 @@ int cap_task_setnice (struct task_struct *p, int nice)return cap_safe_nice (p); } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -/* -* Running a setuid root program raises your capabilities. -* Killing your own setuid root processes was previously -* allowed. -* We must preserve legacy signal behavior in this case. -*/ -if (p->uid == current->uid) -return 0 ; - -if (secid) -/* -* Signal sent as a particular user. -* Capabilities are ignored. May be wrong, but it's the -* only thing we can do at the moment. -* Used only by usb drivers? -*/ -return 0 ; -if (cap_issubset (p->cap_permitted, current->cap_permitted)) -return 0 ; -if (capable (CAP_KILL )) -return 0 ; - -return -EPERM ; -} -/* * called from kernel/sys.c for prctl(PR_CABSET_DROP) * done without task_capability_lock() because it introduces @@ -596,11 +568,6 @@ int cap_task_setnice (struct task_struct *p, int nice) {return 0 ; } -int cap_task_kill (struct task_struct *p,struct siginfo *info, -int sig,u32 secid) -{ -return 0 ; -}#endif int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,security/smack/smack_lsm.c +0
-5@@ -1117,11 +1117,6 @@ static int smack_task_movememory(struct task_struct *p)static int smack_task_kill (struct task_struct *p,struct siginfo *info,int sig,u32 secid) { -int rc; - - rc =cap_task_kill (p, info, sig, secid); -if (rc !=0 ) -return rc;/* * Sending a signal requires that the sender
Quoted 12 lines
* can write the receiver.
Signature · 7 lines
1.5.2.5 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Casey Schaufler
[email protected]