Re: [PATCH] cobalt/posix/syscall: Account for changes in kernel 7.0

On Tue, 2026-04-14 at 17:35 +0200, Jan Kiszka wrote:
> From: Jan Kiszka <[email protected]>
> 
> With 7.0, we no longer have a access to trace print function that takes
> the string size directly. [__]trace_puts will now always invoke strlen
> on the passed string. We therefore need to terminate our string
> properly, even if a larger one was passed.
> 
> We could keep the micro-optimized variant for older kernels, but as that
> is phasing out and the gain is likely minimal (the string is still in L1
> cache after copy_from_user), simply switch all versions to the same
> pattern.
> 
> Signed-off-by: Jan Kiszka <[email protected]>
> ---
>  kernel/cobalt/posix/syscall.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/cobalt/posix/syscall.c b/kernel/cobalt/posix/syscall.c
> index a26da154c4..bc1825676f 100644
> --- a/kernel/cobalt/posix/syscall.c
> +++ b/kernel/cobalt/posix/syscall.c
> @@ -46,6 +46,12 @@
>  #include "../debug.h"
>  #include <trace/events/cobalt-posix.h>
>  
> +#if LINUX_VERSION_CODE >= KERNEL_VERSION(7,0,0)
> +#include <linux/trace_printk.h>
> +#else
> +#include <linux/kernel.h>
> +#endif
> +
>  /* Syscall must run into the Linux domain. */
>  #define __xn_exec_lostage    0x1
>  /* Syscall must run into the Xenomai domain. */
> @@ -194,12 +200,13 @@ static COBALT_SYSCALL(ftrace_puts, current,
>  	char buf[256];
>  	unsigned len;
>  
> -	len = cobalt_strncpy_from_user(buf, str, sizeof(buf));
> +	len = cobalt_strncpy_from_user(buf, str, sizeof(buf) - 1);
>  	if (len < 0)
>  		return -EFAULT;
>  
>  #ifdef CONFIG_TRACING
> -	__trace_puts(_THIS_IP_, buf, len);
> +	buf[sizeof(buf) - 1] = 0;

Shouldn't we terminate at buf[len]?

buf[len+1] to buf[sizeof(buf)-1] might be garbage, no?

> +	trace_puts(buf);
>  #endif
>  
>  	return 0;
> -- 
> 2.47.3