Re: Size of irq field

From: Julien Grall <julien.grall@citrix.com>
To: Iurii Konovalenko <iurii.konovalenko@globallogic.com>,
	Julien Grall <julien.grall@linaro.org>
Cc: Embedded-pv-devel@lists.xenproject.org,
	Stefano Stabellini <Stefano.Stabellini@eu.citrix.com>,
	Ian Campbell <Ian.Campbell@citrix.com>,
	"xen-devel@lists.xen.org" <xen-devel@lists.xen.org>
Subject: Re: Size of irq field
Date: Thu, 2 Apr 2015 16:34:58 +0100	[thread overview]
Message-ID: <551D61A2.6020201@citrix.com> (raw)
In-Reply-To: <CABc08z+1KgYoXtG7TZTzokOVqT42tAN3NKSeYhGFp+sV1JufMg@mail.gmail.com>

(CC Stefano and Ian)

On 02/04/2015 16:05, Iurii Konovalenko wrote:
> Hi, Julien!

Hello Iurii,

>
> During bringing up Xen on Renesas Lager board we faced with problem.
> A lot of Xen sources relies on statement, that IRQ number is less then
> 256 and variables, parameters, fields etc. are of type uint8_t. But we
> can have IRQs, that are greater then 255, for example on RCar H2 SoC.
> Also, as I saw from one of your latest commits, GICv supports 1020
> physical interrupts. As a result, overflow can occur. So it seems
> logical to increase all irq staff to uint16_t or uint32_t. We have
> local patches for increasing some structures, functions etc.
> But before pushing these patches I want to know your opinion, what do
> you think about this problem? What type should we use uint16_t or
> uint32_t? How to find all places where uint8_t type is used for IRQ?

I'm a bit surprised. I looked through the hypervisor and libxl and I was 
able to found only one place (DOMCTL_irq_permission which is not really 
supported on ARM). We are usually using uint32_t/int for IRQ.

Can you give an example of files/structures using uint8_t for IRQ?

Ideally the IRQ should use uint32_t. This is allow us to support LPIs 
(IRQ number start a 8192 up to a very high number).

> Best regards.

Regards,

-- 
Julien Grall