From mboxrd@z Thu Jan  1 00:00:00 1970
From: mark.rutland@arm.com (Mark Rutland)
Date: Mon, 8 Feb 2016 12:03:17 +0000
Subject: [PATCH v4 07/13] arm64: Change cpu_resume() to enable mmu early
 then access sleep_sp by va
In-Reply-To: <20160208115552.GA5471@red-moon>
References: <1453977766-20907-1-git-send-email-james.morse@arm.com>
 <1453977766-20907-8-git-send-email-james.morse@arm.com>
 <20160205162617.GD31547@red-moon> <56B859D9.3080601@arm.com>
 <20160208115552.GA5471@red-moon>
Message-ID: <20160208120317.GA7700@leverpostej>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Mon, Feb 08, 2016 at 11:55:52AM +0000, Lorenzo Pieralisi wrote:
> On Mon, Feb 08, 2016 at 09:03:21AM +0000, James Morse wrote:
> > Hi Lorenzo,
> > 
> > On 05/02/16 16:26, Lorenzo Pieralisi wrote:
> > >> cpu_do_{suspend,resume} could be further reduced to not restore: cpacr_el1,
> > >> mdscr_el1, tcr_el1, vbar_el1 and sctlr_el1, all of which are set by
> > >> __cpu_setup(). However these values all contain res0 bits that may be used
> > >> to enable future features.
> > > 
> > > This patch is a very nice clean-up, a comment below.
> > > 
> > > I think that for registers like tcr_el1 and sctlr_el1 we should define
> > > a restore mask (to avoid overwriting bits set-up by __cpu_setup), eg
> > > current code that restores the tcr_el1 seems wrong to me, see below.
> > 
> > Presumably this should be two masks, one to find RES0 bits that are
> > set, and are assumed to have some new meaning, and another to find
> > RES1 bits that have been cleared.
> 
> For the time being, it is ok to just fix t0sz restore which means
> that either you avoid overwriting tcr_el1.t0sz in cpu_do_resume()
> or you force the t0sz value field to be whatever value is already
> present in the register (ie set-up in __cpu_setup through
> tcr_set_idmap_t0sz) and finally set it to correct the default value
> in __cpu_suspend_exit() using the correct procedure:
> 
> - set reserved ttbr0
> - flush tlb
> - cpu_set_default_tcr_t0sz

You can use cpu_uninstall_idmap() [1] to do that, which Catalin has
queued as part of my pagetable rework [2]. That will also install the
appropriate TTBR0 context for the current thread, if any.

Mark.

[1] https://git.kernel.org/cgit/linux/kernel/git/arm64/linux.git/commit/?h=for-next/pgtable&id=7036610bbd05a5269fa1d25c1c000ad3465c2906
[2] https://git.kernel.org/cgit/linux/kernel/git/arm64/linux.git/log/?h=for-next/pgtable