From: Colin Cross <ccross@android.com>
Subject: Re: [PATCH] ARM: vfp: Always save VFP state in vfp_pm_suspend
Date: Mon, 14 Feb 2011 10:35:37 -0800
Lines: 50
Message-ID: <AANLkTik_Jey_PtRmr530FVckA6RXHESeX+CyoJC=ZTkR@mail.gmail.com>
References: <1297638813-1315-1-git-send-email-ccross@android.com>
	<1297683742.30092.11.camel@e102109-lin.cambridge.arm.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: linux-arm-kernel@lists.infradead.org,
	Russell King <linux@arm.linux.org.uk>,
	linux-kernel@vger.kernel.org
To: Catalin Marinas <catalin.marinas@arm.com>
Envelope-to: glk-linux-kernel-3@lo.gmane.org
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=google.com; s=beta;
        h=mime-version:sender:in-reply-to:references:date
         :x-google-sender-auth:message-id:subject:from:to:cc:content-type
         :content-transfer-encoding;
        b=Lio4jom4+RruCzH/a6zheXvUcTpZvSX7eMk6Ld1+PQQybP02I0+Cv4+PgGac2GoAYX
         tXciu0c/KOGts2phbyMg==
In-Reply-To: <1297683742.30092.11.camel@e102109-lin.cambridge.arm.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk
Archived-At: <http://permalink.gmane.org/gmane.linux.kernel/1099943>

On Mon, Feb 14, 2011 at 3:42 AM, Catalin Marinas
<catalin.marinas@arm.com> wrote:
> On Sun, 2011-02-13 at 23:13 +0000, Colin Cross wrote:
>> vfp_pm_suspend should save the VFP state any time there is
>> a last_VFP_context. =A0If it only saves when the VFP is enabled,
>> the state can get lost when, on a UP system:
>> =A0 =A0Thread 1 uses the VFP
>> =A0 =A0Context switch occurs to thread 2, VFP is disabled but the
>> =A0 =A0 =A0 VFP context is not saved to allow lazy save and restore
>> =A0 =A0Thread 2 initiates suspend
>> =A0 =A0vfp_pm_suspend is called with the VFP disabled, but the
>> =A0 =A0 =A0 context has not been saved.
>
> At this point is it guaranteed that the thread won't migrate to anoth=
er
> CPU? If not, we should use get/put_cpu.

Yes, VFP suspend is implemented with a sysdev, which is suspended
after disable_nonboot_cpus.

>> --- a/arch/arm/vfp/vfpmodule.c
>> +++ b/arch/arm/vfp/vfpmodule.c
>> @@ -415,13 +415,12 @@ static int vfp_pm_suspend(struct sys_device *d=
ev, pm_message_t state)
>> =A0 =A0 =A0 =A0 struct thread_info *ti =3D current_thread_info();
>> =A0 =A0 =A0 =A0 u32 fpexc =3D fmrx(FPEXC);
>>
>> - =A0 =A0 =A0 /* if vfp is on, then save state for resumption */
>> - =A0 =A0 =A0 if (fpexc & FPEXC_EN) {
>> + =A0 =A0 =A0 /* save state for resume */
>> + =A0 =A0 =A0 if (last_VFP_context[ti->cpu]) {
>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 printk(KERN_DEBUG "%s: saving vfp st=
ate\n", __func__);
>> - =A0 =A0 =A0 =A0 =A0 =A0 =A0 vfp_save_state(&ti->vfpstate, fpexc);
>> -
>> - =A0 =A0 =A0 =A0 =A0 =A0 =A0 /* disable, just in case */
>> - =A0 =A0 =A0 =A0 =A0 =A0 =A0 fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 fmxr(FPEXC, fpexc | FPEXC_EN);
>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 vfp_save_state(last_VFP_context[ti->cp=
u], fpexc);
>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 fmxr(FPEXC, fpexc & ~FPEXC_EN);
>> =A0 =A0 =A0 =A0 }
>
> We may want to set the last_VFP_context to NULL so that after resumin=
g
> (to the same thread) we force the VFP reload from the vfpstate
> structure. The vfp_support_entry code ignores the reloading if the
> last_VFP_context is the same as vfpstate.

Right, will fix.