Re: [PATCH v2] x86/syscalls: Mark expected switch fall-throughs

[Dmitry V. Levin]

On Wed, Mar 27, 2019 at 11:52:19PM +0100, Thomas Gleixner wrote:
> On Thu, 28 Mar 2019, Dmitry V. Levin wrote:
> > On Wed, Mar 27, 2019 at 03:29:16PM +0100, Thomas Gleixner wrote:
> > > On Wed, 27 Mar 2019, Dmitry V. Levin wrote:
> > > > On Tue, Mar 26, 2019 at 04:12:45PM +0100, Oleg Nesterov wrote:
> > > > > On 03/23, Thomas Gleixner wrote:
> > > > [...]
> > > > > >  2) syscall_set_arguments() has been introduced in 2008 and we still have
> > > > > >     no caller. Instead of polishing it, can it be removed completely or are
> > > > > >     there plans to actually use it?
> > > > > 
> > > > > I think it can die.
> > > > 
> > > > When PTRACE_GET_SYSCALL_INFO is finally squeezed into the kernel,
> > > > we could discuss adding PTRACE_SET_SYSCALL_INFO as well, and it
> > > > will need syscall_set_arguments().
> > > 
> > > So if that ever happens, then adding the code back isn't rocket
> > > science. But if not, then there is no point in carrying the dead horse
> > > around another 11 years.
> > 
> > Given that it took me roughly 4 months to get a relatively simple revert
> > of commit 5e937a9ae913 accepted into linux-next, adding the code back
> > might be time-consuming.
> > 
> > Could we delay the removal of syscall_set_arguments() until
> > PTRACE_GET_SYSCALL_INFO is merged into the kernel?
> > I hope it won't take another 11 years.
> 
> Hope dies last :)
> 
> Seriously. If we keep it can we at least remove all the unused arguments
> which we have on both functions to simplify the whole mess?

In case of syscall_set_arguments() I think we can safely remove
"i" and "n" arguments assuming i == 0 and n == 6.

All I can say about syscall_get_arguments() is that
- all current users invoke it with i == 0,
- all current users that invoke it with n != 6 are in kernel/trace/trace_syscalls.c
so it may actually be invoked with n < 6.


-- 
ldv

Attachment: signature.asc
Description: PGP signature