netdev - Re: [PATCH v3 bpf-next 01/10] treewide: remove struct-pass-by-value from tracepoints arguments

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20180322164831.01519df6@gandalf.local.home>
Date:   Thu, 22 Mar 2018 16:48:31 -0400
From:   Steven Rostedt <rostedt@...dmis.org>
To:     Alexei Starovoitov <ast@...com>
Cc:     <davem@...emloft.net>, <daniel@...earbox.net>,
        <torvalds@...ux-foundation.org>, <peterz@...radead.org>,
        <netdev@...r.kernel.org>, <kernel-team@...com>,
        <linux-api@...r.kernel.org>
Subject: Re: [PATCH v3 bpf-next 01/10] treewide: remove struct-pass-by-value
 from tracepoints arguments

On Thu, 22 Mar 2018 12:31:12 -0700
Alexei Starovoitov <ast@...com> wrote:

> On 3/22/18 11:11 AM, Steven Rostedt wrote:
> > On Thu, 22 Mar 2018 11:01:48 -0700
> > Alexei Starovoitov <ast@...com> wrote:
> >  
> >> From: Alexei Starovoitov <ast@...nel.org>
> >>
> >> Fix all tracepoint arguments to pass structures (large and small) by reference
> >> instead of by value.
> >> Avoiding passing large structs by value is a good coding style.
> >> Passing small structs sometimes is beneficial, but in all cases
> >> it makes no difference vs readability of the code.
> >> The subsequent patch enforces that all tracepoints args are either integers
> >> or pointers and fit into 64-bit.  
> >
> > But some of these structures are used to force type checking, and are
> > just the same size as a number. That's why they don't have "struct" in
> > front of them. Like pmd_t. Will the subsequent patches really break if
> > the structure itself has one element that is of size long? Just seems
> > to add extra code to pass in an address to something that fits into a
> > single register.  
> 
> yeah. C doesn't allow casting of 'struct s { u64 var };' into u64
> without massive hacks and aliasing warnings by compiler.
> CAST_TO_U64 macro in patch 7 will prevent tracepoint arguments to be
> things like pmd_t. It's not perfect, but doing & of pmd_t variable
> is imo clean enough as you can see in this patch.
> The macro can be tweaked to do the cast like
> *(sizeof(typeof(arg))*)&arg,
> but there is no way to get rid of compiler warning.

OK, but instead of changing it to pass by reference, why not just pass
the value. That is:

 static void xen_set_pte_atomic(pte_t *ptep, pte_t pte)
 {
-	trace_xen_mmu_set_pte_atomic(ptep, pte);
+	trace_xen_mmu_set_pte_atomic(ptep, native_pte_val(pte));
 	set_64bit((u64 *)ptep, native_pte_val(pte));
 }

It shouldn't add any extra code, as those helper functions are
basically just special casts.

-- Steve