Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:3 as permitted sender) client-ip=2620:137:e000::3:3;
Message-ID: <8bf013fa-0fa8-4ebc-9776-0cd6c134972e@rivosinc.com>
Date:   Thu, 14 Sep 2023 19:42:21 +0200
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] tracing/user_events: align uaddr on unsigned long
 alignment
Content-Language: en-US
To:     Beau Belgrave <beaub@linux.microsoft.com>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org
References: <20230914131102.179100-1-cleger@rivosinc.com>
 <20230914164218.GA450-beaub@linux.microsoft.com>
From:   =?UTF-8?B?Q2zDqW1lbnQgTMOpZ2Vy?= <cleger@rivosinc.com>
In-Reply-To: <20230914164218.GA450-beaub@linux.microsoft.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 14/09/2023 18:42, Beau Belgrave wrote:
> On Thu, Sep 14, 2023 at 03:11:02PM +0200, Clément Léger wrote:
>> enabler->uaddr can be aligned on 32 or 64 bits. If aligned on 32 bits,
>> this will result in a misaligned access on 64 bits architectures since
>> set_bit()/clear_bit() are expecting an unsigned long (aligned) pointer.
>> On architecture that do not support misaligned access, this will crash
>> the kernel. Align uaddr on unsigned long size to avoid such behavior.
>> This bug was found while running kselftests on RISC-V.
>>
>> Fixes: 7235759084a4 ("tracing/user_events: Use remote writes for event enablement")
>> Signed-off-by: Clément Léger <cleger@rivosinc.com>
> 
> Thanks for fixing! I have a few comments on this.
> 
> I unfortunately do not have RISC-V hardware to validate this on.
> 
>> ---
>>  kernel/trace/trace_events_user.c | 12 +++++++++---
>>  1 file changed, 9 insertions(+), 3 deletions(-)
>>
>> diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
>> index 6f046650e527..580c0fe4b23e 100644
>> --- a/kernel/trace/trace_events_user.c
>> +++ b/kernel/trace/trace_events_user.c
>> @@ -479,7 +479,7 @@ static int user_event_enabler_write(struct user_event_mm *mm,
>>  				    bool fixup_fault, int *attempt)
>>  {
>>  	unsigned long uaddr = enabler->addr;
>> -	unsigned long *ptr;
>> +	unsigned long *ptr, bit_offset;
>>  	struct page *page;
>>  	void *kaddr;
>>  	int ret;
>> @@ -511,13 +511,19 @@ static int user_event_enabler_write(struct user_event_mm *mm,
>>  	}
>>  
>>  	kaddr = kmap_local_page(page);
>> +
>> +	bit_offset = uaddr & (sizeof(unsigned long) - 1);
>> +	if (bit_offset) {
>> +		bit_offset *= 8;
> 
> I think for future readers of this code it would be more clear to use
> BITS_PER_BYTE instead of the hardcoded 8. Given we always align on a
> "natural" boundary, I believe the bit_offset will always be 32 bits.
> 
> A comment here might help clarify why we do this as well in case folks
> don't see the change description.

Hi Beau,

Yes sure, I'll add a comment and use the define as well.

> 
>> +		uaddr &= ~(sizeof(unsigned long) - 1);
> 
> Shouldn't this uaddr change be done before calling pin_user_pages_remote()
> to ensure things cannot go bad? (I don't think they can, but it looks a
> little odd).

Indeed, I don't think that will cause any problem since pin_user_pages
will return a page aligned address anyway and that aligning uaddr will
not yield any page crossing. But I'll check to be sure and move that
before the call if needed.

Clément

> 
> Thanks,
> -Beau
> 
>> +	}
>>  	ptr = kaddr + (uaddr & ~PAGE_MASK);
>>  
>>  	/* Update bit atomically, user tracers must be atomic as well */
>>  	if (enabler->event && enabler->event->status)
>> -		set_bit(ENABLE_BIT(enabler), ptr);
>> +		set_bit(ENABLE_BIT(enabler) + bit_offset, ptr);
>>  	else
>> -		clear_bit(ENABLE_BIT(enabler), ptr);
>> +		clear_bit(ENABLE_BIT(enabler) + bit_offset, ptr);
>>  
>>  	kunmap_local(kaddr);
>>  	unpin_user_pages_dirty_lock(&page, 1, true);
>> -- 
>> 2.40.1