Skip to contentSkip to conversation

posix-timers: Require privilege to enable timer ID restore mode

PATCH In review Ashutosh Desai 2 participants 3 messages Apr 9 – Apr 14 Posix clocks and timers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 posix-timers: Require privilege to enable timer ID restore mode

PR_TIMER_CREATE_RESTORE_IDS_ON switches timer_create() into a mode where the caller supplies explicit timer IDs, intended exclusively for CRIU checkpoint/restore. The UAPI comment is explicit: "Don't use for normal operations as the result might be undefined."

Despite that, the prctl handler has no capability check. Any unprivileged process can enable the mode. Every comparable CRIU prctl in the kernel gates on checkpoint_restore_ns_capable() - PR_SET_MM_MAP, PR_SET_MM_EXE_FILE, etc. This one should too.

Add the check to the ON case only. OFF and GET are left unrestricted so a process can always query or clear the flag without privilege.

Cc Thomas Gleixner <[email protected]>
Cc Anna-Maria Behnsen <[email protected]>
---
 kernel/time/posix-timers.c | 3 +++
 1 file changed, 3 insertions(+)
kernel/time/posix-timers.c +3 -0
index 413e2389f..b8582416b 100644
@@ -9,6 +9,7 @@
  *
  * These are all the functions necessary to implement POSIX clocks & timers
  */
+#include <linux/capability.h>
 #include <linux/compat.h>
 #include <linux/compiler.h>
 #include <linux/init.h>
@@ -380,6 +381,8 @@ long posixtimer_create_prctl(unsigned long ctrl)
 		current->signal->timer_create_restore_ids = 0;
 		return 0;
 	case PR_TIMER_CREATE_RESTORE_IDS_ON:
+		if (!checkpoint_restore_ns_capable(current_user_ns()))
+			return -EPERM;
 		current->signal->timer_create_restore_ids = 1;
 		return 0;
 	case PR_TIMER_CREATE_RESTORE_IDS_GET:
Signature · 1 line
2.34.1
#1 lore raw
Quoted 10 lines · Ashutosh Desai

PR_TIMER_CREATE_RESTORE_IDS_ON switches timer_create() into a mode where the caller supplies explicit timer IDs, intended exclusively for CRIU checkpoint/restore. The UAPI comment is explicit: "Don't use for normal operations as the result might be undefined."

Despite that, the prctl handler has no capability check. Any unprivileged process can enable the mode. Every comparable CRIU prctl in the kernel gates on checkpoint_restore_ns_capable() - PR_SET_MM_MAP, PR_SET_MM_EXE_FILE, etc. This one should too.

This statement is entirely accurate. Capability checks are typically required only when the operation has security implications or allows a process to bypass standard resource isolation.

Quoted 36 lines

Add the check to the ON case only. OFF and GET are left unrestricted so a process can always query or clear the flag without privilege.

Cc Thomas Gleixner <[email protected]>
Cc Anna-Maria Behnsen <[email protected]>
---
 kernel/time/posix-timers.c | 3 +++
 1 file changed, 3 insertions(+)
kernel/time/posix-timers.c +3 -1
index 413e2389f..b8582416b 100644
@@ -9,6 +9,7 @@
  *
  * These are all the functions necessary to implement POSIX clocks & timers
  */
+#include <linux/capability.h>
 #include <linux/compat.h>
 #include <linux/compiler.h>
 #include <linux/init.h>
@@ -380,6 +381,8 @@ long posixtimer_create_prctl(unsigned long ctrl)
                current->signal->timer_create_restore_ids = 0;
                return 0;
        case PR_TIMER_CREATE_RESTORE_IDS_ON:
+               if (!checkpoint_restore_ns_capable(current_user_ns()))
+                       return -EPERM;
                current->signal->timer_create_restore_ids = 1;
                return 0;
        case PR_TIMER_CREATE_RESTORE_IDS_GET:
--

2.34.1

#2 lore raw

Thanks Andrei. Looking at the introducing commit (ec2d0c0), I see Thomas explicitly considered the rogue userspace scenario and concluded it is harmless. I will drop this patch.

Ashutosh