x86: Address various objtool complaints
Subsystems (3)
Peter Zijlstra
maintainercover x86: Address various objtool complaints
Courtesy of 0-day:
fs/ntfs3/ntfs3.prelink.o: warning: objtool: ni_read_frame() falls through to next function ni_readpage_cmpr.cold()
vmlinux.o: warning: objtool: enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_enter_from_user_mode+0x28: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_enter_from_user_mode_prepare+0x24: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: irqentry_enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section
vmlinux.o: warning: objtool: exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool: irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text section
Peter Zijlstra
maintainerPATCH 1/3 objtool: Mark __ubsan_handle_builtin_unreachable() as noreturn
fs/ntfs3/ntfs3.prelink.o: warning: objtool: ni_read_frame() falls through to next function ni_readpage_cmpr.cold()
That is in fact:
000000000000124a <ni_read_frame.cold>:
124a: 44 89 e0 mov %r12d,%eax
124d: 0f b6 55 98 movzbl -0x68(%rbp),%edx
1251: 48 c7 c7 00 00 00 00 mov $0x0,%rdi 1254: R_X86_64_32S .data+0x1380
1258: 48 89 c6 mov %rax,%rsi
125b: e8 00 00 00 00 call 1260 <ni_read_frame.cold+0x16> 125c: R_X86_64_PLT32 __ubsan_handle_shift_out_of_bounds-0x4
1260: 48 8d 7d cc lea -0x34(%rbp),%rdi
1264: e8 00 00 00 00 call 1269 <ni_read_frame.cold+0x1f> 1265: R_X86_64_PLT32 __tsan_read4-0x4
1269: 8b 45 cc mov -0x34(%rbp),%eax
126c: e9 00 00 00 00 jmp 1271 <ni_read_frame.cold+0x27> 126d: R_X86_64_PC32 .text+0x19109
1271: 48 8b 75 a0 mov -0x60(%rbp),%rsi
1275: 48 63 d0 movslq %eax,%rdx
1278: 48 c7 c7 00 00 00 00 mov $0x0,%rdi 127b: R_X86_64_32S .data+0x13a0
127f: 89 45 88 mov %eax,-0x78(%rbp)
1282: e8 00 00 00 00 call 1287 <ni_read_frame.cold+0x3d> 1283: R_X86_64_PLT32 __ubsan_handle_shift_out_of_bounds-0x4
1287: 8b 45 88 mov -0x78(%rbp),%eax
128a: e9 00 00 00 00 jmp 128f <ni_read_frame.cold+0x45> 128b: R_X86_64_PC32 .text+0x19098
128f: 48 c7 c7 00 00 00 00 mov $0x0,%rdi 1292: R_X86_64_32S .data+0x11f0
1296: e8 00 00 00 00 call 129b <ni_readpage_cmpr.cold> 1297: R_X86_64_PLT32 __ubsan_handle_builtin_unreachable-0x4000000000000129b <ni_readpage_cmpr.cold>:
Tell objtool that __ubsan_handle_builtin_unreachable() is a noreturn.
--- tools/objtool/check.c | 1 + 1 file changed, 1 insertion(+)
tools/objtool/check.c +1 -0
@@ -184,6 +184,7 @@ static bool __dead_end_function(struct o"do_group_exit" ,"stop_this_cpu" ,"__invalid_creds" , + "__ubsan_handle_builtin_unreachable", };if (!func)
Peter Zijlstra
maintainerPATCH 2/3 x86/cpu: Elide KCSAN for cpu_has() and friends
In reply to Peter Zijlstravmlinux.o: warning: objtool: enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_enter_from_user_mode+0x28: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_enter_from_user_mode_prepare+0x24: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: irqentry_enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section
--- arch/x86/include/asm/cpufeature.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/include/asm/cpufeature.h +1 -1
@@ -51,7 +51,7 @@ extern const char * const x86_power_flagextern const char *const x86_bug_flags[NBUGINTS *32 ];#define test_cpu_cap (c, bit) \ - test_bit(bit, (unsigned long *)((c)->x86_capability)) + arch_test_bit(bit, (unsigned long *)((c)->x86_capability))/* * There are 32 bits/features in each mask word. The high bits
Re: [PATCH 2/3] x86/cpu: Elide KCSAN for cpu_has() and friends
Quoted 7 lines · Peter Zijlstra
vmlinux.o: warning: objtool:
enter_from_user_mode+0x24: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_enter_from_user_mode+0x28: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_enter_from_user_mode_prepare+0x24: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_enter_from_user_mode+0x24: call to__kcsan_check_access()leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>
And should fix the same issue for other sanitizers that get instrumented test_bit() via asm-generic/bitops/instrumented-*.h, not just KCSAN.
Quoted 17 lines
--- arch/x86/include/asm/cpufeature.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/x86/include/asm/cpufeature.h +1
-1@@ -51,7 +51,7 @@ extern const char * const x86_power_flagextern const char *const x86_bug_flags[NBUGINTS *32 ];#define test_cpu_cap (c, bit) \ - test_bit(bit, (unsigned long *)((c)->x86_capability)) + arch_test_bit(bit, (unsigned long *)((c)->x86_capability))/* * There are 32 bits/features in each mask word. The high bits
Re: [PATCH 2/3] x86/cpu: Elide KCSAN for cpu_has() and friends
In reply to Peter ZijlstraQuoted 7 lines · Peter Zijlstra
vmlinux.o: warning: objtool:
enter_from_user_mode+0x24: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_enter_from_user_mode+0x28: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_enter_from_user_mode_prepare+0x24: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_enter_from_user_mode+0x24: call to__kcsan_check_access()leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>
An explanation about *why* this fixes it would help, as I have no idea from looking at the patch.
Quoted 17 lines
--- arch/x86/include/asm/cpufeature.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/x86/include/asm/cpufeature.h +1
-1@@ -51,7 +51,7 @@ extern const char * const x86_power_flagextern const char *const x86_bug_flags[NBUGINTS *32 ];#define test_cpu_cap (c, bit) \ - test_bit(bit, (unsigned long *)((c)->x86_capability)) + arch_test_bit(bit, (unsigned long *)((c)->x86_capability))/* * There are 32 bits/features in each mask word. The high bits
Signature · 1 line
Josh
Re: [PATCH 2/3] x86/cpu: Elide KCSAN for cpu_has() and friends
Quoted 11 lines · Josh Poimboeuf
Quoted 7 lines · Peter Zijlstra
vmlinux.o: warning: objtool:
enter_from_user_mode+0x24: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_enter_from_user_mode+0x28: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_enter_from_user_mode_prepare+0x24: call to__kcsan_check_access()leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_enter_from_user_mode+0x24: call to__kcsan_check_access()leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>An explanation about *why* this fixes it would help, as I have no idea from looking at the patch.
How about something like:
| As x86 uses the <asm-generic/bitops/instrumented-*.h> headers, the | regular forms of all bitops are instrumented with explicit calls to | KASAN and KCSAN checks. As these are explicit calls, these are not | suppressed by the noinstr function attribute. | | This can result in calls to those check functions in noinstr code, which | objtool warns about: | | vmlinux.o: warning: objtool: enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section | vmlinux.o: warning: objtool: syscall_enter_from_user_mode+0x28: call to __kcsan_check_access() leaves .noinstr.text section | vmlinux.o: warning: objtool: syscall_enter_from_user_mode_prepare+0x24: call to __kcsan_check_access() leaves .noinstr.text section | vmlinux.o: warning: objtool: irqentry_enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section | | Prevent this by using the arch_*() bitops, which are the underlying | bitops without explciit instrumentation.
Thanks,
Mark.
Quoted 22 lines
Quoted 18 lines
--- arch/x86/include/asm/cpufeature.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/x86/include/asm/cpufeature.h +1
-1@@ -51,7 +51,7 @@ extern const char * const x86_power_flagextern const char *const x86_bug_flags[NBUGINTS *32 ];#define test_cpu_cap (c, bit) \ - test_bit(bit, (unsigned long *)((c)->x86_capability)) + arch_test_bit(bit, (unsigned long *)((c)->x86_capability))/* * There are 32 bits/features in each mask word. The high bits Signature · 1 line
Josh
[tip: objtool/urgent] x86/cpu: Elide KCSAN for cpu_has() and friends
In reply to Peter ZijlstraThe following commit has been merged into the objtool/urgent branch of tip:
The following commit has been merged into the objtool/urgent branch of tip:
Commit-ID: a6a5eb269f6f3a2fe392f725a8d9052190c731e2 Gitweb: https://git.kernel.org/tip/a6a5eb269f6f3a2fe392f725a8d9052190c731e2 Author: Peter Zijlstra <[email protected]> AuthorDate: Mon, 02 May 2022 12:15:23 +02:00 Committer: Peter Zijlstra <[email protected]> CommitterDate: Fri, 27 May 2022 12:34:43 +02:00
x86/cpu: Elide KCSAN for cpu_has() and friends
As x86 uses the <asm-generic/bitops/instrumented-*.h> headers, the regular forms of all bitops are instrumented with explicit calls to KASAN and KCSAN checks. As these are explicit calls, these are not suppressed by the noinstr function attribute.
This can result in calls to those check functions in noinstr code, which objtool warns about:
vmlinux.o: warning: objtool: enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_enter_from_user_mode+0x28: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_enter_from_user_mode_prepare+0x24: call to __kcsan_check_access() leaves .noinstr.text section vmlinux.o: warning: objtool: irqentry_enter_from_user_mode+0x24: call to __kcsan_check_access() leaves .noinstr.text section
Prevent this by using the arch_*() bitops, which are the underlying bitops without explciit instrumentation.
--- arch/x86/include/asm/cpufeature.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/include/asm/cpufeature.h +1 -1
@@ -54,7 +54,7 @@ extern const char * const x86_power_flags[32];extern const char *const x86_bug_flags[NBUGINTS *32 ];#define test_cpu_cap (c, bit) \ - test_bit(bit, (unsigned long *)((c)->x86_capability)) + arch_test_bit(bit, (unsigned long *)((c)->x86_capability))/* * There are 32 bits/features in each mask word. The high bits
Peter Zijlstra
maintainerPATCH 3/3 jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
In reply to Peter ZijlstraWhen building x86_64 with JUMP_LABEL=n it's possible for instrumentation to sneak into noinstr:
vmlinux.o: warning: objtool: exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool: irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text section
--- include/linux/jump_label.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/jump_label.h +2 -2
@@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke#include <linux/atomic.h> #include <linux/bug.h> -static inline int static_key_count (struct static_key *key) +static __always_inline int static_key_count(struct static_key *key) { -return atomic_read (&key->enabled); +return READ_ONCE_NOCHECK (&key->enabled.count); } static __always_inline void jump_label_init(void)
Re: [PATCH 3/3] jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
Quoted 17 lines · Peter Zijlstra
When building
x86_64with JUMP_LABEL=n it's possible for instrumentation to sneak into noinstr:vmlinux.o: warning: objtool:
exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>--- include/linux/jump_label.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/jump_label.h extern void static_key_disable_cpuslocke · line 262
Quoted 9 lines
include/linux/jump_label.h +2
-2@@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke#include <linux/atomic.h> #include <linux/bug.h> -static inline int static_key_count (struct static_key *key) +static __always_inline int static_key_count(struct static_key *key) { -return atomic_read (&key->enabled); +return READ_ONCE_NOCHECK (&key->enabled.count); }
Did arch_atomic_read() not work? Or should this also be unchecked in instrumented functions?
Peter Zijlstra
maintainerRe: [PATCH 3/3] jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
Quoted 21 lines · Marco Elver
Quoted 20 lines · Peter Zijlstra
When building
x86_64with JUMP_LABEL=n it's possible for instrumentation to sneak into noinstr:vmlinux.o: warning: objtool:
exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>--- include/linux/jump_label.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)include/linux/jump_label.h +0
-0@@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke#include <linux/atomic.h> #include <linux/bug.h>
Quoted 9 lines
include/linux/jump_label.h extern void static_key_disable_cpuslocke · line 262
Quoted 6 lines
include/linux/jump_label.h +2
-2-static inline int static_key_count (struct static_key *key) +static __always_inline int static_key_count(struct static_key *key) { -return atomic_read (&key->enabled); +return READ_ONCE_NOCHECK (&key->enabled.count); }Did
arch_atomic_read()not work? Or should this also be unchecked in instrumented functions?
Urgh, also I think the above is broken for me forgetting to remove the &.
I think arch_atomic_read() should work, lemme try.
Peter Zijlstra
maintainerPATCH v2 3/3 jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
In reply to Peter ZijlstraSubject: jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
From: Peter Zijlstra <[email protected]>
Date: Mon May 2 12:30:20 CEST 2022
When building x86_64 with JUMP_LABEL=n it's possible for instrumentation to sneak into noinstr:
vmlinux.o: warning: objtool: exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool: syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool: irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text section
--- include/linux/jump_label.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/jump_label.h +2 -2
@@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke#include <linux/atomic.h> #include <linux/bug.h> -static inline int static_key_count (struct static_key *key) +static __always_inline int static_key_count(struct static_key *key) { -return atomic_read (&key->enabled); +return arch_atomic_read (&key->enabled.count); } static __always_inline void jump_label_init(void)
Re: [PATCH v2 3/3] jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
Quoted 21 lines · Peter Zijlstra
Subject: jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
From: Peter Zijlstra <[email protected]>
Date: Mon May 2 12:30:20 CEST 2022When building
x86_64with JUMP_LABEL=n it's possible for instrumentation to sneak into noinstr:vmlinux.o: warning: objtool:
exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>--- include/linux/jump_label.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/jump_label.h extern void static_key_disable_cpuslocke · line 261
Quoted 8 lines
include/linux/jump_label.h +2
-2@@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke#include <linux/atomic.h> #include <linux/bug.h> -static inline int static_key_count (struct static_key *key) +static __always_inline int static_key_count(struct static_key *key) { -return atomic_read (&key->enabled); +return arch_atomic_read (&key->enabled.count);
Curious if this compiles - s/.count// ?
Quoted 3 lines
include/linux/jump_label.h +0
-0} static __always_inline void jump_label_init(void)
Peter Zijlstra
maintainerRe: [PATCH v2 3/3] jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
Quoted 25 lines · Marco Elver
Quoted 24 lines · Peter Zijlstra
Subject: jump_label,noinstr: Avoid instrumentation for JUMP_LABEL=n builds
From: Peter Zijlstra <[email protected]>
Date: Mon May 2 12:30:20 CEST 2022When building
x86_64with JUMP_LABEL=n it's possible for instrumentation to sneak into noinstr:vmlinux.o: warning: objtool:
exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section vmlinux.o: warning: objtool:irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text sectionReported-by kernel test robot <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>--- include/linux/jump_label.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)include/linux/jump_label.h +0
-0@@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke#include <linux/atomic.h> #include <linux/bug.h>
Quoted 7 lines
include/linux/jump_label.h extern void static_key_disable_cpuslocke · line 261
Quoted 5 lines
include/linux/jump_label.h +2
-2-static inline int static_key_count (struct static_key *key) +static __always_inline int static_key_count(struct static_key *key) { -return atomic_read (&key->enabled); +return arch_atomic_read (&key->enabled.count);Curious if this compiles - s/.count// ?
It does if you have JUMP_LABEL=y... (-:
*sigh*... how I about I go back to mowing the lawn and try again later...