Skip to contentSkip to conversation

ARM: kasan: Fix __get_user_check failure with kasan

PATCH Superseded Lexi Shao 4 participants 10 messages Aug 25, 2021 – Sep 20, 2021 Arm (32-bit)

lore
Share & tools
lore

Lexi Shao

author
#0 lore raw

PATCH 1/1 ARM: kasan: Fix __get_user_check failure with kasan

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +3 -1
index a13d90206472..a6eb9af74870 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__e = __tmp_e;						\
 	})
 
 #define get_user(x, p)							\
Signature · 7 lines
2.12.3


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
#1 lore raw

25.08.2021 09:46, Lexi Shao пишет:

Quoted 70 lines

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +1 -0
index a13d90206472..a6eb9af74870 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 237

Quoted 10 lines
arch/arm/include/asm/uaccess.h +2 -1
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__e = __tmp_e;						\
 	})
 
 #define get_user(x, p)							\

I successfully loaded KDE Plasma with this fix, which was impossible previously. Thank you!

Is it guaranteed that r2 register won't be clobbered as well?

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

author
#2 lore raw
Quoted 79 lines

25.08.2021 09:46, Lexi Shao пишет:

Quoted 78 lines

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +3 -1
index a13d90206472..a6eb9af74870 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__e = __tmp_e;						\
 	})
 
Quoted 7 lines

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 237

Quoted 2 lines
arch/arm/include/asm/uaccess.h +0 -0
 #define get_user(x, p)							\

I successfully loaded KDE Plasma with this fix, which was impossible previously. Thank you!

Is it guaranteed that r2 register won't be clobbered as well?

Value of r2 is saved before calling __asan_store4:

c01f6dbc:       e5951008        ldr     r1, [r5, #8]
c01f6dc0:       e1a00008        mov     r0, r8
c01f6dc4:       e2411001        sub     r1, r1, #1
c01f6dc8:       eb0b020e        bl      c04b7608 <__get_user_4>
c01f6dcc:       e1a00007        mov     r0, r7
// move value of r2 to r5
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
// the assignment, saving r5(orig r2) to dest
c01f6dd8:       e5875000        str     r5, [r7]
c01f6ddc:       e1a05000        mov     r5, r0

I can't explain it clearly why the compiler save r2 but not r0 though...

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Reviewed-by on v1 1/1
#3 lore raw
In reply to Lexi Shao
Quoted 46 lines · Lexi Shao

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

Quoted 6 lines

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")

---

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#4 lore raw
In reply to Lexi Shao
Quoted 39 lines · Lexi Shao

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Quoted 2 lines

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

This description doesn't actually make it clear why this is failing in this case, until one looks at get_futex_value_locked() and realises that what is actually going on here is:

	*dest = (typeof(*(p))) __r2;

which is why we end up with the store there.

Quoted 26 lines

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +1 -0
index a13d90206472..a6eb9af74870 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 234

Quoted 8 lines
arch/arm/include/asm/uaccess.h +2 -1
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__e = __tmp_e;						\

There is no need to re-assign __tmp_e back to __e - you can just return __tmp_e here.

Signature · 7 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#5 lore raw
Quoted 80 lines · Lexi Shao
Quoted 39 lines

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Quoted 2 lines

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

This description doesn't actually make it clear why this is failing in this case, until one looks at get_futex_value_locked() and realises that what is actually going on here is:

	*dest = (typeof(*(p))) __r2;

which is why we end up with the store there.

Quoted 29 lines

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +1 -0
index a13d90206472..a6eb9af74870 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\
 		default: __e = __get_user_bad(); break;			\
 		}							\
Quoted 8 lines

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 234

Quoted 5 lines
arch/arm/include/asm/uaccess.h +2 -1
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__e = __tmp_e;						\

There is no need to re-assign __tmp_e back to __e - you can just return __tmp_e here.

Yes you are right, will send the updated patch in the next email.

Lexi

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

author
#6 lore raw

PATCH v2 1/1 ARM: kasan: Fix __get_user_check failure with kasan

In reply to Russell King (Oracle)

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

This line in function get_futex_value_locked

	ret = __get_user(*dest, from);
is expanded to
	*dest = (typeof(*(p))) __r2; ,

in macro __get_user_check. Writing to pointer dest triggers kasan check and overwrites the return value of __get_user_x function. The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +3 -1
index a13d90206472..d9db752c51fe 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__tmp_e;						\
 	})
 
 #define get_user(x, p)							\
Signature · 7 lines
2.12.3


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Tested-by on v2 1/1
#7 lore raw

28.08.2021 05:25, Lexi Shao пишет:

Quoted 76 lines

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

This line in function get_futex_value_locked

	ret = __get_user(*dest, from);
is expanded to
	*dest = (typeof(*(p))) __r2; ,

in macro __get_user_check. Writing to pointer dest triggers kasan check and overwrites the return value of __get_user_x function. The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +1 -0
index a13d90206472..d9db752c51fe 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 237

Quoted 10 lines
arch/arm/include/asm/uaccess.h +2 -1
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__tmp_e;						\
 	})
 
 #define get_user(x, p)							\

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#8 lore raw

31.08.2021 12:51, Dmitry Osipenko пишет:

Quoted 85 lines

28.08.2021 05:25, Lexi Shao пишет:

Quoted 84 lines

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

This line in function get_futex_value_locked

	ret = __get_user(*dest, from);
is expanded to
	*dest = (typeof(*(p))) __r2; ,

in macro __get_user_check. Writing to pointer dest triggers kasan check and overwrites the return value of __get_user_x function. The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
---
 arch/arm/include/asm/uaccess.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +3 -1
index a13d90206472..d9db752c51fe 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
 		register unsigned long __l asm("r1") = __limit;		\
 		register int __e asm("r0");				\
 		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
 		switch (sizeof(*(__p))) {				\
 		case 1:							\
 			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
 			break;						\
 		default: __e = __get_user_bad(); break;			\
 		}							\
+		__tmp_e = __e;						\
 		uaccess_restore(__ua_flags);				\
 		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__tmp_e;						\
 	})
 
Quoted 5 lines

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 237

Quoted 2 lines
arch/arm/include/asm/uaccess.h +0 -0
 #define get_user(x, p)							\

Note that you may need to submit this patch to
https://www.armlinux.org.uk/developer/patches/ to speed up the applying.

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Reviewed-by on v2 1/1
#9 lore raw
In reply to Lexi Shao
Quoted 52 lines · Lexi Shao

In macro __get_user_check defined in arch/arm/include/asm/uaccess.h, error code is store in register int __e(r0). When kasan is enabled, assigning value to kernel address might trigger kasan check, which unexpectedly overwrites r0 and causes undefined behavior on arm kasan images.

One example is failure in do_futex and results in process soft lockup. Log:

watchdog: BUG: soft lockup - CPU#0 stuck for 62946ms! [rs:main

Q:Reg:1151]
...
(__asan_store4) from (futex_wait_setup+0xf8/0x2b4) (futex_wait_setup) from (futex_wait+0x138/0x394) (futex_wait) from (do_futex+0x164/0xe40) (do_futex) from (sys_futex_time32+0x178/0x230) (sys_futex_time32) from (ret_fast_syscall+0x0/0x50)

The soft lockup happens in function futex_wait_setup. The reason is function get_futex_value_locked always return EINVAL, thus pc jump back to retry label and causes looping.

This line in function get_futex_value_locked

	ret = __get_user(*dest, from);
is expanded to
	*dest = (typeof(*(p))) __r2; ,

in macro __get_user_check. Writing to pointer dest triggers kasan check and overwrites the return value of __get_user_x function. The assembly code of get_futex_value_locked in kernel/futex.c:
...
c01f6dc8: eb0b020e bl c04b7608 <__get_user_4>
// "x = (typeof(*(p))) __r2;" triggers kasan check and r0 is overwritten

c01f6dcc:       e1a00007        mov     r0, r7
c01f6dd0:       e1a05002        mov     r5, r2
c01f6dd4:       eb04f1e6        bl      c0333574 <__asan_store4>
c01f6dd8:       e5875000        str     r5, [r7]

// save ret value of __get_user(*dest, from), which is dest address now
c01f6ddc: e1a05000 mov r5, r0
...
// checking return value of __get_user failed
c01f6e00: e3550000 cmp r5, #0
...
c01f6e0c: 01a00005 moveq r0, r5
// assign return value to EINVAL
c01f6e10: 13e0000d mvnne r0, #13

Return value is the destination address of get_user thus certainly non-zero, so get_futex_value_locked always return EINVAL.

Fix it by using a tmp vairable to store the error code before the assignment. This fix has no effects to non-kasan images thanks to compiler optimization. It only affects cases that overwrite r0 due to kasan check.

Quoted 5 lines

This should fix bug discussed in link:
[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Fixes 421015713b30 ("ARM: 9017/2: Enable KASan for ARM")
Quoted 18 lines
---
  arch/arm/include/asm/uaccess.h | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +1 -0
index a13d90206472..d9db752c51fe 100644
@@ -200,6 +200,7 @@ extern int __get_user_64t_4(void *);
  		register unsigned long __l asm("r1") = __limit;		\
  		register int __e asm("r0");				\
  		unsigned int __ua_flags = uaccess_save_and_enable();	\
+		int __tmp_e;						\
  		switch (sizeof(*(__p))) {				\
  		case 1:							\
  			if (sizeof((x)) >= 8)				\
@@ -227,9 +228,10 @@ extern int __get_user_64t_4(void *);
  			break;						\

arch/arm/include/asm/uaccess.h extern int __get_user_64t_4(void *); · line 237

Quoted 10 lines
arch/arm/include/asm/uaccess.h +2 -1
  		default: __e = __get_user_bad(); break;			\
  		}							\
+		__tmp_e = __e;						\
  		uaccess_restore(__ua_flags);				\
  		x = (typeof(*(p))) __r2;				\
-		__e;							\
+		__tmp_e;						\
  	})
 
  #define get_user(x, p)							\

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel