2022-06-27 09:13:01

by Li kunyu

[permalink] [raw]
Subject: [PATCH] tools: Strong conversion of void type pointer could be removed

The void pointer argument does not require a cast assignment because it
is the address passed.

Signed-off-by: Li kunyu <[email protected]>
---
tools/testing/selftests/x86/fsgsbase.c | 2 +-
tools/testing/selftests/x86/test_vsyscall.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/x86/fsgsbase.c b/tools/testing/selftests/x86/fsgsbase.c
index 8c780cce941d..5d99261317e1 100644
--- a/tools/testing/selftests/x86/fsgsbase.c
+++ b/tools/testing/selftests/x86/fsgsbase.c
@@ -63,7 +63,7 @@ static void clearhandler(int sig)

static void sigsegv(int sig, siginfo_t *si, void *ctx_void)
{
- ucontext_t *ctx = (ucontext_t*)ctx_void;
+ ucontext_t *ctx = ctx_void;

if (!want_segv) {
clearhandler(SIGSEGV);
diff --git a/tools/testing/selftests/x86/test_vsyscall.c b/tools/testing/selftests/x86/test_vsyscall.c
index 5b45e6986aea..2416941a0952 100644
--- a/tools/testing/selftests/x86/test_vsyscall.c
+++ b/tools/testing/selftests/x86/test_vsyscall.c
@@ -184,7 +184,7 @@ static volatile unsigned long segv_err;

static void sigsegv(int sig, siginfo_t *info, void *ctx_void)
{
- ucontext_t *ctx = (ucontext_t *)ctx_void;
+ ucontext_t *ctx = ctx_void;

segv_err = ctx->uc_mcontext.gregs[REG_ERR];
siglongjmp(jmpbuf, 1);
--
2.18.2


2022-06-27 20:27:42

by Shuah Khan

[permalink] [raw]
Subject: Re: [PATCH] tools: Strong conversion of void type pointer could be removed

On 6/27/22 2:56 AM, Li kunyu wrote:
> The void pointer argument does not require a cast assignment because it
> is the address passed.
>

Please include information on you found this problem with output
from the tool if any used. Send v2 with that information included
in the commit log.

> Signed-off-by: Li kunyu <[email protected]>
> ---
> tools/testing/selftests/x86/fsgsbase.c | 2 +-
> tools/testing/selftests/x86/test_vsyscall.c | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/x86/fsgsbase.c b/tools/testing/selftests/x86/fsgsbase.c
> index 8c780cce941d..5d99261317e1 100644
> --- a/tools/testing/selftests/x86/fsgsbase.c
> +++ b/tools/testing/selftests/x86/fsgsbase.c
> @@ -63,7 +63,7 @@ static void clearhandler(int sig)
>
> static void sigsegv(int sig, siginfo_t *si, void *ctx_void)
> {
> - ucontext_t *ctx = (ucontext_t*)ctx_void;
> + ucontext_t *ctx = ctx_void;
>
> if (!want_segv) {
> clearhandler(SIGSEGV);
> diff --git a/tools/testing/selftests/x86/test_vsyscall.c b/tools/testing/selftests/x86/test_vsyscall.c
> index 5b45e6986aea..2416941a0952 100644
> --- a/tools/testing/selftests/x86/test_vsyscall.c
> +++ b/tools/testing/selftests/x86/test_vsyscall.c
> @@ -184,7 +184,7 @@ static volatile unsigned long segv_err;
>
> static void sigsegv(int sig, siginfo_t *info, void *ctx_void)
> {
> - ucontext_t *ctx = (ucontext_t *)ctx_void;
> + ucontext_t *ctx = ctx_void;
>
> segv_err = ctx->uc_mcontext.gregs[REG_ERR];
> siglongjmp(jmpbuf, 1);
>

thanks,
-- Shuah

2022-06-29 03:09:05

by Li kunyu

[permalink] [raw]
Subject: Re: [PATCH] tools: Strong conversion of void type pointer could be removed


Hi Shuah, now I can't paste the test code, so I could write a demo and paste it:


-------------source---------------

#include <stdio.h>
#include <stdlib.h>

struct ucontext {
struct ucontext *uc_link;
unsigned long uc_flags;
sigset_t uc_sigmask;
struct ucontext *uc_mcontext;
};
typedef struct ucontext ucontext_t;

void sigsegv(void *ctx_void)
{
ucontext_t *ctx = (ucontext_t*)ctx_void;
ucontext_t *ctx2 = (int *)ctx_void;
ucontext_t *ctx3 = ctx_void;
printf("ctx:%p, ctx2:%p, ctx3:%p.\n", ctx, ctx2, ctx3);
}

int main() {
ucontext_t *test = malloc(sizeof(ucontext_t));
sigsegv(test);
return 0;
}

--------------------------------------

The result is CTX: 0x563D96CE5010, CTX2:0x563D96CE5010, CTx3:0x563D96CE5010.
Now force ucontext_t and int pointers are the same as the addresses obtained without forced conversion.

Now I'll paste the assembly code for them:


|0x700 <sigsegv> push %rbp │
│0x701 <sigsegv+1> mov %rsp,%rbp │
│0x704 <sigsegv+4> sub $0x30,%rsp │
│0x708 <sigsegv+8> mov %rdi,-0x28(%rbp) │
│0x70c <sigsegv+12> mov -0x28(%rbp),%rax │
│0x710 <sigsegv+16> mov %rax,-0x8(%rbp) │
│0x714 <sigsegv+20> mov -0x28(%rbp),%rax │
│0x718 <sigsegv+24> mov %rax,-0x10(%rbp) │
│0x71c <sigsegv+28> mov -0x28(%rbp),%rax │
│0x720 <sigsegv+32> mov %rax,-0x18(%rbp) │
│0x724 <sigsegv+36> mov -0x18(%rbp),%rcx │
│0x728 <sigsegv+40> mov -0x10(%rbp),%rdx │
│0x72c <sigsegv+44> mov -0x8(%rbp),%rax │
│0x730 <sigsegv+48> mov %rax,%rsi │
│0x733 <sigsegv+51> lea 0xba(%rip),%rdi # 0x7f4 │
│0x73a <sigsegv+58> mov $0x0,%eax │
│0x73f <sigsegv+63> callq 0x5a0 <printf@plt> │
│0x744 <sigsegv+68> nop │
│0x745 <sigsegv+69> leaveq │
│0x746 <sigsegv+70> retq

2022-06-30 00:25:45

by Shuah Khan

[permalink] [raw]
Subject: Re: [PATCH] tools: Strong conversion of void type pointer could be removed

On 6/28/22 8:42 PM, Li kunyu wrote:
>
> Hi Shuah, now I can't paste the test code, so I could write a demo and paste it:
>
>

What is the "test code"?

How did you find the problem? Which tool did you use it?

thanks,
-- Shuah