linux-kernel - Re: [PATCH v1 2/6] mfd: cros

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

Message-ID: <56BCA031.3010702@collabora.com>
Date:	Thu, 11 Feb 2016 15:52:33 +0100
From:	Tomeu Vizoso <tomeu.vizoso@...labora.com>
To:	gwendal@...omium.org
Cc:	Linux Kernel <linux-kernel@...r.kernel.org>,
	Sameer Nanda <snanda@...omium.org>,
	Benson Leung <bleung@...omium.org>,
	Enric Balletbò <enric.balletbo@...labora.co.uk>,
	Vic Yang <victoryang@...omium.org>,
	Vincent Palatin <vpalatin@...omium.org>,
	Randall Spangler <rspangler@...omium.org>,
	Gwendal Grignou <gwendal@...omium.org>,
	Vic Yang <victoryang@...gle.com>,
	Olof Johansson <olof@...om.net>, linux-input@...r.kernel.org,
	Dmitry Torokhov <dmitry.torokhov@...il.com>,
	Lee Jones <lee.jones@...aro.org>
Subject: Re: [PATCH v1 2/6] mfd: cros_ec: Add MKBP event support

On 02/05/2016 06:24 PM, Gwendal Grignou wrote:
> We should not used kmalloc when get events, these functions are called
> quite often.
> Gwendal.
> 
> On Fri, Feb 5, 2016 at 5:32 AM, Tomeu Vizoso <tomeu.vizoso@...labora.com
> <mailto:tomeu.vizoso@...labora.com>> wrote:
> 
>     +static int cros_ec_get_host_command_version_mask(struct
>     cros_ec_device *ec_dev,
>     +       u16 cmd, u32 *mask)
>     +{
>     +       struct ec_params_get_cmd_versions *pver;
>     +       struct ec_response_get_cmd_versions *rver;
>     +       struct cros_ec_command *msg;
>     +       int ret;
>     +
>     +       msg = kmalloc(sizeof(*msg) + max(sizeof(rver), sizeof(pver)),
>     +                     GFP_KERNEL);
>     +       if (!msg)
>     +               return -ENOMEM;
> 
>  Victor's version in https://chromium-review.googlesource.com/272954
> looks cleaner: no malloc, no need to cast rver.

I agree that it looks cleaner, but how would you allocate the payload at
build time if it has to be max(sizeof(*pver), sizeof(*rver))?

>     +static int get_next_event(struct cros_ec_device *ec_dev)
>     +{
>     +       struct cros_ec_command *msg;
>     +       int ret;
>     +
>     +       msg = kmalloc(sizeof(*msg) + sizeof(ec_dev->event_data),
>     GFP_KERNEL);
>     +       if (!msg)
>     +               return -ENOMEM;
> 
> Same grip about malloc(). This function will be called very often when
> sensors are used.

Ok.

>     +static int get_keyboard_state_event(struct cros_ec_device *ec_dev)
>     +{
>     +       struct cros_ec_command *msg;
>     +
>     +       msg = kmalloc(sizeof(*msg) + sizeof(ec_dev->event_data.data),
>     +                     GFP_KERNEL);
>     +       if (!msg)
>     +               return -ENOMEM;
> 
> Given this command will be used on every key press, it is better to
> pre-allocate data (like
> in https://chromium-review.googlesource.com/276768) and use msg on the
> stack like Victor's changes.

Ok.

Thanks,

Tomeu