linux-kernel - Re: [PATCH] arm64/cpufeature: Convert hook_lock to raw_spin_lock_t in cpu_enable

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

Date:   Thu, 30 May 2019 14:55:05 +0100
From:   Julien Grall <julien.grall@....com>
To:     Will Deacon <will.deacon@....com>
Cc:     linux-kernel@...r.kernel.org, linux-rt-users@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, tglx@...utronix.de,
        rostedt@...dmis.org, bigeasy@...utronix.de, suzuki.poulose@....com,
        catalin.marinas@....com, dave.martin@....com
Subject: Re: [PATCH] arm64/cpufeature: Convert hook_lock to raw_spin_lock_t in
 cpu_enable_ssbs()

Hi Will,

On 5/30/19 1:01 PM, Will Deacon wrote:
> On Thu, May 30, 2019 at 12:30:58PM +0100, Julien Grall wrote:
>> cpu_enable_ssbs() is called via stop_machine() as part of the cpu_enable
>> callback. A spin lock is used to ensure the hook is registered before
>> the rest of the callback is executed.
>>
>> On -RT spin_lock() may sleep. However, all the callees in stop_machine()
>> are expected to not sleep. Therefore a raw_spin_lock() is required here.
>>
>> Given this is already done under stop_machine() and the work done under
>> the lock is quite small, the latency should not increase too much.
>>
>> Signed-off-by: Julien Grall <julien.grall@....com>
>>
>> ---
>>
>> It was noticed when looking at the current use of spin_lock in
>> arch/arm64. I don't have a platform calling that callback, so I have
>> hacked the code to reproduce the error and check it is now fixed.
>> ---
>>   arch/arm64/kernel/cpufeature.c | 6 +++---
>>   1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
>> index ca27e08e3d8a..2a7159fda3ce 100644
>> --- a/arch/arm64/kernel/cpufeature.c
>> +++ b/arch/arm64/kernel/cpufeature.c
>> @@ -1194,14 +1194,14 @@ static struct undef_hook ssbs_emulation_hook = {
>>   static void cpu_enable_ssbs(const struct arm64_cpu_capabilities *__unused)
>>   {
>>   	static bool undef_hook_registered = false;
>> -	static DEFINE_SPINLOCK(hook_lock);
>> +	static DEFINE_RAW_SPINLOCK(hook_lock);
>>   
>> -	spin_lock(&hook_lock);
>> +	raw_spin_lock(&hook_lock);
>>   	if (!undef_hook_registered) {
>>   		register_undef_hook(&ssbs_emulation_hook);
>>   		undef_hook_registered = true;
>>   	}
>> -	spin_unlock(&hook_lock);
>> +	raw_spin_unlock(&hook_lock);
> 
> Makes sense to me. We could probably avoid the lock entirely if we wanted
> to (via atomic_dec_if_positive), but I'm not sure it's really worth it.

I would prefer to remove the lock if it is possible. However, I was 
under the impression the lock is necessary so the hook is registered 
before any CPU attempt to configure the PSTATE.

Cheers,

-- 
Julien Grall