linux-kernel - Re: [PATCH] x86/traps: use conditional_{cli,sti} in preempt_conditinal_{cli

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:	Sat, 16 Jan 2016 14:46:31 -0800
From:	Andy Lutomirski <luto@...capital.net>
To:	Borislav Petkov <bp@...e.de>
Cc:	Alexander Kuleshov <kuleshovmail@...il.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Ingo Molnar <mingo@...hat.com>,
	"H . Peter Anvin" <hpa@...or.com>, X86 ML <x86@...nel.org>,
	Andy Lutomirski <luto@...nel.org>,
	Dave Hansen <dave.hansen@...ux.intel.com>,
	Wang Nan <wangnan0@...wei.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] x86/traps: use conditional_{cli,sti} in preempt_conditinal_{cli_sti}

On Sat, Jan 16, 2016 at 11:06 AM, Borislav Petkov <bp@...e.de> wrote:
> On Sat, Jan 16, 2016 at 06:58:07PM +0600, Alexander Kuleshov wrote:
>> The 3d2a71a596bd9 commit (x86, traps: converge do_debug handlers by
>> Alexander van Heukelum <heukelum@...tmail.fm>) introduces two functions:
>> preempt_conditional_sti/cli() which are enables/disables interrupts
>> depends on state of the interrupt enable flag and increments/decrements
>> the preempt counter.
>>
>> In the same time arch/x86/kernel/traps.c defines two similar inline
>> functions: conditional_{sti,cli} which are do the same, but without
>> touch of the preempt counter. Let's use these functions in the
>> preempt_conditional_{sti,cli} instead of duplication of 'if' statemets.
>>
>> Signed-off-by: Alexander Kuleshov <kuleshovmail@...il.com>
>> ---
>>  arch/x86/kernel/traps.c | 6 ++----
>>  1 file changed, 2 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
>> index ade185a..30ec8fa 100644
>> --- a/arch/x86/kernel/traps.c
>> +++ b/arch/x86/kernel/traps.c
>> @@ -92,8 +92,7 @@ static inline void conditional_sti(struct pt_regs *regs)
>>  static inline void preempt_conditional_sti(struct pt_regs *regs)
>>  {
>>       preempt_count_inc();
>> -     if (regs->flags & X86_EFLAGS_IF)
>> -             local_irq_enable();
>> +     conditional_sti(regs);
>
> What I would do is kill both preempt_conditional_sti() and
> preempt_conditional_cli() instead. Why?
>
> Because call sites become more readable:
>
>         preempt_disable();
>
>         if (regs->flags & X86_EFLAGS_IF)
>                 local_irq_enable();
>
> and
>
>         if (regs->flags & X86_EFLAGS_IF)
>                 local_irq_disable();
>
>         preempt_enable_no_resched();
>
> Those preempt_* variants are just silly.
>
> Yes, I'd delete conditional_cli() because nothing uses it. And since I'm
> deleting crap, I'd delete conditional_sti() too because it's naming is
> ugly. You have to know that "STI" means Set Interrupt Flag. Now, if it
> were called
>
>         cond_local_irq_enable()
>
> that would be better.
>
> So yeah, IMO, the most understandable variant would be:
>
>         preempt_disable();
>         cond_local_irq_enable();
>
> and the opposing pair:
>
>         cond_local_irq_disable();
>         preempt_enable_no_resched();
>
> And *that* is actually understandable at a quick glance.
>
> But that's just me.
>

I agree.  Every time I look at those helpers I wonder what the condition is.

Also, "preempt_conditional_sti" meaning "enable interrupts if they
were enabled but *don't* enable preemption" just seems backwards.

--Andy