syscall trace at kernel land

[Dave Hylands]

Hi Rajat,

On Tue, Jan 11, 2011 at 9:26 PM, Rajat Sharma <fs.rajat at gmail.com> wrote:
> Hi Dave,
>
> My understanding was based out of linux/errno.h. Maybe the below
> comment is in context of glibc abstracting out ERESTART* error codes.
>
> #ifdef __KERNEL__
>
> /*
>  * These should never be seen by user programs.  To return one of ERESTART*
>  * codes, signal_pending() MUST be set.  Note that ptrace can observe these
>  * at syscall exit tracing, but they will never be left for the debugged user
>  * process to see.
>  */
> #define ERESTARTSYS     512
> #define ERESTARTNOINTR  513
> #define ERESTARTNOHAND  514     /* restart if no handler.. */
> #define ENOIOCTLCMD     515     /* No ioctl command */
> #define ERESTART_RESTARTBLOCK 516 /* restart by calling sys_restart_syscall */

Yeah - that's consistent with my understanding. The code flow would be
something like this

user-mode program calls ioctl
    ioctl does some stuff
    signal happens
    ioctl detects signal (typically by a call returning -EINTR) and
decides to return -ERESTARTSYS
    sys handler detects pending signal
        user-mode signal handler is called
    sys handler re-issues ioctl
    ioctl does some stuff
    ioctl returns normally
user mode program sees normal reutrn code

So ioctl returns -ERESTARTSYS, but the user mode program is completely
oblivious to the fact that this happened.

glibc doesn't even get to know that the -ERESTARTSYS code was returned
by ioctl. It gets intercepted and processed by the kernel.

Dave Hylands