Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 27 Apr 2021 19:05:56 -0700
From:   Kees Cook <keescook@chromium.org>
To:     Andy Lutomirski <luto@amacapital.net>
Cc:     "H. Peter Anvin" <hpa@zytor.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Ingo Molnar <mingo@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Andrew Lutomirski <luto@kernel.org>,
        Borislav Petkov <bp@alien8.de>, linux-kernel@vger.kernel.org,
        oleg@redhat.com, Will Drewry <wad@chromium.org>
Subject: Re: pt_regs->ax == -ENOSYS
Message-ID: <202104271857.4708C9A25@keescook>
References: <202104271619.0DBE456@keescook>
 <78327AF2-575F-415E-98D4-6ECDFE311D0F@amacapital.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <78327AF2-575F-415E-98D4-6ECDFE311D0F@amacapital.net>
Precedence: bulk

On Tue, Apr 27, 2021 at 04:51:06PM -0700, Andy Lutomirski wrote:
> Fortunately there is not, and never will be, a syscall -1.  But I
> agree that calling max syscall + 1 should behave identically to calling
> a nonexistent syscall in the middle of the table.

If that happens, we have to separate the meaning of -1L from ptrace,
seccomp, etc. (i.e. we can't just add an "else { result = -ENOSYS; }" to
the syscall table dispatching code, since that'll overwrite any written
return value when the syscall is meant to be skipped with a specific
return value set by ptrace/seccomp.

syscall_trace_enter() will currently return either -1 or the
syscall. Which means someone making a "syscall -1" will get the skip
semantics currently (though the preloaded -ENOSYS results in the
"expected" outcome).

arm64 recently had to untangle this too:

15956689a0e6 arm64: compat: Ensure upper 32 bits of x0 are zero on syscall return
59ee987ea47c arm64: ptrace: Add a comment describing our syscall entry/exit trap ABI
139dbe5d8ed3 arm64: syscall: Expand the comment about ptrace and syscall(-1)
d83ee6e3e75d arm64: ptrace: Use NO_SYSCALL instead of -1 in syscall_trace_enter()

-- 
Kees Cook