[PATCH] ia32: strncpy does not null terminate string

[PATCH] ia32: strncpy does not null terminate string

[Roel Kluin]

With `sizeof(lastcomm) - 1` strncpy() will null terminate the string.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
---
To test this:

#include <stdio.h>
#include <string.h>

char a[10];
char b[10];

int main()
{
        const char* str = "0123456789012";
        strncpy(a, str, sizeof(a));
        strncpy(b, str, sizeof(b) - 1);
        printf("String a was %s, b was %s\n", a, b);

        return 0;
}

Output:
String a was 0123456789012345678, b was 012345678

diff --git a/arch/x86/ia32/sys_ia32.c b/arch/x86/ia32/sys_ia32.c
index 085a8c3..b114f57 100644
--- a/arch/x86/ia32/sys_ia32.c
+++ b/arch/x86/ia32/sys_ia32.c
@@ -687,7 +687,7 @@ long sys32_vm86_warning(void)
 		compat_printk(KERN_INFO
 			      "%s: vm86 mode not supported on 64 bit kernel\n",
 			      me->comm);
-		strncpy(lastcomm, me->comm, sizeof(lastcomm));
+		strncpy(lastcomm, me->comm, sizeof(lastcomm) - 1);
 	}
 	return -ENOSYS;
 }

Re: [PATCH] ia32: strncpy does not null terminate string

[Roel Kluin]

strlcpy() will always null terminate the string.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
---
> The proof is flawed: Global variables are automatically '\0'
> initialized.

Hmm, I see, how about using strlcpy instead?

diff --git a/arch/x86/ia32/sys_ia32.c b/arch/x86/ia32/sys_ia32.c
index 085a8c3..528972f 100644
--- a/arch/x86/ia32/sys_ia32.c
+++ b/arch/x86/ia32/sys_ia32.c
@@ -687,7 +687,7 @@ long sys32_vm86_warning(void)
 		compat_printk(KERN_INFO
 			      "%s: vm86 mode not supported on 64 bit kernel\n",
 			      me->comm);
-		strncpy(lastcomm, me->comm, sizeof(lastcomm));
+		strlcpy(lastcomm, me->comm, sizeof(lastcomm));
 	}
 	return -ENOSYS;
 }

Re: [PATCH] ia32: strncpy does not null terminate string

[H. Peter Anvin]

Roel Kluin wrote:
> 
> diff --git a/arch/x86/ia32/sys_ia32.c b/arch/x86/ia32/sys_ia32.c
> index 085a8c3..b114f57 100644
> --- a/arch/x86/ia32/sys_ia32.c
> +++ b/arch/x86/ia32/sys_ia32.c
> @@ -687,7 +687,7 @@ long sys32_vm86_warning(void)
>  		compat_printk(KERN_INFO
>  			      "%s: vm86 mode not supported on 64 bit kernel\n",
>  			      me->comm);
> -		strncpy(lastcomm, me->comm, sizeof(lastcomm));
> +		strncpy(lastcomm, me->comm, sizeof(lastcomm) - 1);
>  	}
>  	return -ENOSYS;
>  }

I don't see any problem with the code as written.  It is of course
correct that strncpy() doesn't null-terminate (it null-pads, which is
somewhat inefficient, but has info leak advantages), *however*, the only
user (it's a local static variable) uses strncpy(), so that's fine.

It probably could be done cleaner, but there is no actual bug, so I
personally don't see any reason to change it just in the interest of
avoiding churn.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.