linux-kernel - Re: [PATCH 02/10] irqchip/gic-v3-its: Consolidate LPI_PENDBASE

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: <f37b69a8-64fd-490d-1680-9f6aef235e10@arm.com>
Date:   Mon, 24 Sep 2018 11:54:02 +0100
From:   Suzuki K Poulose <suzuki.poulose@....com>
To:     Julien Thierry <julien.thierry@....com>,
        Marc Zyngier <marc.zyngier@....com>,
        linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Cc:     Ard Biesheuvel <ard.biesheuvel@...aro.org>,
        Jeremy Linton <jeremy.linton@....com>,
        Jeffrey Hugo <jhugo@...eaurora.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Jason Cooper <jason@...edaemon.net>
Subject: Re: [PATCH 02/10] irqchip/gic-v3-its: Consolidate LPI_PENDBASE_SZ
 usage



On 24/09/18 11:50, Julien Thierry wrote:
> Hi,
> 
> On 24/09/18 11:33, Suzuki K Poulose wrote:
>> Hi Marc,
>>
>> On 21/09/18 20:59, Marc Zyngier wrote:
>>> LPI_PENDING_SZ is always used in conjunction with a max(). Let's
>>> factor this in the definition of the macro, and simplify the rest
>>> of the code.
>>>
>>> Signed-off-by: Marc Zyngier <marc.zyngier@....com>
>>> ---
>>>   drivers/irqchip/irq-gic-v3-its.c | 12 ++++--------
>>>   1 file changed, 4 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/drivers/irqchip/irq-gic-v3-its.c 
>>> b/drivers/irqchip/irq-gic-v3-its.c
>>> index c2df341ff6fa..ed6aab11e019 100644
>>> --- a/drivers/irqchip/irq-gic-v3-its.c
>>> +++ b/drivers/irqchip/irq-gic-v3-its.c
>>> @@ -62,7 +62,7 @@ static u32 lpi_id_bits;
>>>    */
>>>   #define LPI_NRBITS        lpi_id_bits
>>>   #define LPI_PROPBASE_SZ        ALIGN(BIT(LPI_NRBITS), SZ_64K)
>>> -#define LPI_PENDBASE_SZ        ALIGN(BIT(LPI_NRBITS) / 8, SZ_64K)
>>> +#define LPI_PENDBASE_SZ        max_t(u32, SZ_64K, 
>>> ALIGN(BIT(LPI_NRBITS) / 8, SZ_64K))
>>
>> minor nit: The ALIGN() already aligns the given value up to the required
>> alignment. So, if the LPI_NRBITS is guaranteed to be non-zero,
>> we could simply drop the max_t().
>>
> 
> Hmmm, Doesn't ALIGN only aligns down? So if "BIT(LPI_NR_BITS) / 8 < 
> SZ_64K" (i.e. LPI_NRBITS < 20) The ALIGN(..., SZ_64K) would give 0.

Isn't it the ALIGN_DOWN(), which aligns it down ? Following the kernel
definitions :
linux/kernel.h -> uapi/linux/kernel.h
ALIGN(x,a) => 	__ALIGN_KERNEL(x, a)
		\ => __ALIGN_KERNEL_MASK(x, (a -1)
		\ => (((x + (a - 1)) & ~ (a - 1))

Cheers
Suzuki


> 
> Thanks,
> 
>> Rest looks good to me.
>>
>> Suzuki
>>
>>>   #define LPI_PROP_DEFAULT_PRIO    0xa0
>>> @@ -1924,12 +1924,9 @@ static int its_alloc_collections(struct 
>>> its_node *its)
>>>   static struct page *its_allocate_pending_table(gfp_t gfp_flags)
>>>   {
>>>       struct page *pend_page;
>>> -    /*
>>> -     * The pending pages have to be at least 64kB aligned,
>>> -     * hence the 'max(LPI_PENDBASE_SZ, SZ_64K)' below.
>>> -     */
>>> +
>>>       pend_page = alloc_pages(gfp_flags | __GFP_ZERO,
>>> -                get_order(max_t(u32, LPI_PENDBASE_SZ, SZ_64K)));
>>> +                get_order(LPI_PENDBASE_SZ));
>>>       if (!pend_page)
>>>           return NULL;
>>> @@ -1941,8 +1938,7 @@ static struct page 
>>> *its_allocate_pending_table(gfp_t gfp_flags)
>>>   static void its_free_pending_table(struct page *pt)
>>>   {
>>> -    free_pages((unsigned long)page_address(pt),
>>> -           get_order(max_t(u32, LPI_PENDBASE_SZ, SZ_64K)));
>>> +    free_pages((unsigned long)page_address(pt), 
>>> get_order(LPI_PENDBASE_SZ));
>>>   }
>>>   static void its_cpu_init_lpis(void)
>>>
>