linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
From: catalin.marinas@arm.com (Catalin Marinas)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH 03/16] ARM: b.L: introduce helpers for platform coherency exit/setup
Date: Mon, 14 Jan 2013 21:34:53 +0000	[thread overview]
Message-ID: <20130114213450.GA99454@MacBook-Pro.local> (raw)
In-Reply-To: <20130114181006.GB1967@linaro.org>

On Mon, Jan 14, 2013 at 06:10:06PM +0000, Dave Martin wrote:
> On Mon, Jan 14, 2013 at 05:15:28PM +0000, Catalin Marinas wrote:
> > On Mon, Jan 14, 2013 at 05:08:51PM +0000, Dave Martin wrote:
> > > From b64f305c90e7ea585992df2d710f62ec6a7b5395 Mon Sep 17 00:00:00 2001
> > > From: Dave Martin <dave.martin@linaro.org>
> > > Date: Mon, 14 Jan 2013 16:25:47 +0000
> > > Subject: [PATCH] ARM: b.L: Fix outer cache handling for coherency setup/exit helpers
> > > 
> > > This patch addresses the following issues:
> > > 
> > >   * When invalidating stale data from the cache before a read,
> > >     outer caches must be invalidated _before_ inner caches, not
> > >     after, otherwise stale data may be re-filled from outer to
> > >     inner after the inner cache is flushed.
> > > 
> > >     We still retain an inner clean before touching the outer cache,
> > >     to avoid stale data being rewritten from there into the outer
> > >     cache after the outer cache is flushed.
> > > 
> > >   * All the sync_mem() calls synchronise either reads or writes,
> > >     but not both.  This patch splits sync_mem() into separate
> > >     functions for reads and writes, to avoid excessive inner
> > >     flushes in the write case.
> > > 
> > >     The two functions are different from the original sync_mem(),
> > >     to fix the above issues.
> > > 
> > > Signed-off-by: Dave Martin <dave.martin@linaro.org>
> > > ---
> > > NOTE: This patch is build-tested only.
> > > 
> > >  arch/arm/common/bL_entry.c |   57 ++++++++++++++++++++++++++++++++++----------
> > >  1 files changed, 44 insertions(+), 13 deletions(-)
> > > 
> > > diff --git a/arch/arm/common/bL_entry.c b/arch/arm/common/bL_entry.c
> > > index 1ea4ec9..3e1a404 100644
> > > --- a/arch/arm/common/bL_entry.c
> > > +++ b/arch/arm/common/bL_entry.c
> > > @@ -119,16 +119,47 @@ int bL_cpu_powered_up(void)
> > >  
> > >  struct bL_sync_struct bL_sync;
> > >  
> > > -static void __sync_range(volatile void *p, size_t size)
> > > +/*
> > > + * Ensure preceding writes to *p by this CPU are visible to
> > > + * subsequent reads by other CPUs:
> > > + */
> > > +static void __sync_range_w(volatile void *p, size_t size)
> > >  {
> > >  	char *_p = (char *)p;
> > >  
> > >  	__cpuc_flush_dcache_area(_p, size);
> > > -	outer_flush_range(__pa(_p), __pa(_p + size));
> > > +	outer_clean_range(__pa(_p), __pa(_p + size));
> > >  	outer_sync();
> > 
> > It's not part of your patch but I thought about commenting here. The
> > outer_clean_range() already has a cache_sync() operation, so no need for
> > the additional outer_sync().
> > 
> > >  }
> > >  
> > > -#define sync_mem(ptr) __sync_range(ptr, sizeof *(ptr))
> > > +/*
> > > + * Ensure preceding writes to *p by other CPUs are visible to
> > > + * subsequent reads by this CPU:
> > > + */
> > > +static void __sync_range_r(volatile void *p, size_t size)
> > > +{
> > > +	char *_p = (char *)p;
> > > +
> > > +#ifdef CONFIG_OUTER_CACHE
> > > +	if (outer_cache.flush_range) {
> > > +		/*
> > > +		 * Ensure ditry data migrated from other CPUs into our cache
> > > +		 * are cleaned out safely before the outer cache is cleaned:
> > > +		 */
> > > +		__cpuc_flush_dcache_area(_p, size);
> > > +
> > > +		/* Clean and invalidate stale data for *p from outer ... */
> > > +		outer_flush_range(__pa(_p), __pa(_p + size));
> > > +		outer_sync();
> > 
> > Same here.
> 
> Ah, right.  I've seen code do this in various places, and just copy-
> pasted it under the assumption that it is needed.  Our discussion abouto
> ensuring that outer_sync() really does guarantee completion of its
> effects on return still applies.
> 
> Are there any situations when outer_sync() should be called explicitly?

outer_sync() on its own ensures the draining of the PL310 write buffer.
DSB drains the CPU write buffers but PL310 doesn't detect it, so a
separate outer_sync() is needed. In general this is required when you
write a Normal Non-cacheable buffer (but bufferable, e.g. DMA coherent)
and you want to ensure data visibility (DSB+outer_sync(), that's what
the mb() macro does).

-- 
Catalin

  reply	other threads:[~2013-01-14 21:34 UTC|newest]

Thread overview: 132+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-01-10  0:20 [PATCH 00/16] big.LITTLE low-level CPU and cluster power management Nicolas Pitre
2013-01-10  0:20 ` [PATCH 01/16] ARM: b.L: secondary kernel entry code Nicolas Pitre
2013-01-10  7:12   ` Stephen Boyd
2013-01-10 15:30     ` Nicolas Pitre
2013-01-10 15:34   ` Catalin Marinas
2013-01-10 16:47     ` Nicolas Pitre
2013-01-11 11:45       ` Catalin Marinas
2013-01-11 12:05         ` Lorenzo Pieralisi
2013-01-11 12:19         ` Dave Martin
2013-01-10 23:05   ` Will Deacon
2013-01-11  1:26     ` Nicolas Pitre
2013-01-11 10:55       ` Will Deacon
2013-01-11 11:35         ` Dave Martin
2013-01-11 17:16   ` Santosh Shilimkar
2013-01-11 18:10     ` Nicolas Pitre
2013-01-11 18:30       ` Santosh Shilimkar
2013-03-07  7:37   ` Pavel Machek
2013-03-07  8:57     ` Nicolas Pitre
2013-01-10  0:20 ` [PATCH 02/16] ARM: b.L: introduce the CPU/cluster power API Nicolas Pitre
2013-01-10 23:08   ` Will Deacon
2013-01-11  2:30     ` Nicolas Pitre
2013-01-11 10:58       ` Will Deacon
2013-01-11 11:29       ` Dave Martin
2013-01-11 17:26   ` Santosh Shilimkar
2013-01-11 18:33     ` Nicolas Pitre
2013-01-11 18:41       ` Santosh Shilimkar
2013-01-11 19:54         ` Nicolas Pitre
2013-01-10  0:20 ` [PATCH 03/16] ARM: b.L: introduce helpers for platform coherency exit/setup Nicolas Pitre
2013-01-10 12:01   ` Dave Martin
2013-01-10 19:04     ` Nicolas Pitre
2013-01-11 11:30       ` Dave Martin
2013-01-10 16:53   ` Catalin Marinas
2013-01-10 17:59     ` Nicolas Pitre
2013-01-10 21:50       ` Catalin Marinas
2013-01-10 22:31         ` Nicolas Pitre
2013-01-11 10:36           ` Dave Martin
2013-01-10 22:32     ` Nicolas Pitre
2013-01-10 23:13   ` Will Deacon
2013-01-11  1:50     ` Nicolas Pitre
2013-01-11 11:09       ` Dave Martin
2013-01-11 17:46   ` Santosh Shilimkar
2013-01-11 18:07     ` Dave Martin
2013-01-11 18:34       ` Santosh Shilimkar
2013-01-14 17:08   ` Dave Martin
2013-01-14 17:15     ` Catalin Marinas
2013-01-14 18:10       ` Dave Martin
2013-01-14 21:34         ` Catalin Marinas [this message]
2013-01-10  0:20 ` [PATCH 04/16] ARM: b.L: Add baremetal voting mutexes Nicolas Pitre
2013-01-10 23:18   ` Will Deacon
2013-01-11  3:15     ` Nicolas Pitre
2013-01-11 11:03       ` Will Deacon
2013-01-11 16:57       ` Dave Martin
2013-01-10  0:20 ` [PATCH 05/16] ARM: bL_head: vlock-based first man election Nicolas Pitre
2013-01-10  0:20 ` [PATCH 06/16] ARM: b.L: generic SMP secondary bringup and hotplug support Nicolas Pitre
2013-01-11 18:02   ` Santosh Shilimkar
2013-01-14 18:05     ` Achin Gupta
2013-01-15  6:32       ` Santosh Shilimkar
2013-01-15 11:18         ` Achin Gupta
2013-01-15 11:26           ` Santosh Shilimkar
2013-01-15 18:53           ` Dave Martin
2013-01-14 16:35   ` Will Deacon
2013-01-14 16:51     ` Nicolas Pitre
2013-01-15 19:09       ` Dave Martin
2013-01-10  0:20 ` [PATCH 07/16] ARM: bL_platsmp.c: close the kernel entry gate before hot-unplugging a CPU Nicolas Pitre
2013-01-14 16:37   ` Will Deacon
2013-01-14 16:53     ` Nicolas Pitre
2013-01-14 17:00       ` Will Deacon
2013-01-14 17:11         ` Catalin Marinas
2013-01-14 17:15         ` Nicolas Pitre
2013-01-14 17:23           ` Will Deacon
2013-01-14 18:26           ` Russell King - ARM Linux
2013-01-14 18:49             ` Nicolas Pitre
2013-01-15 18:40             ` Dave Martin
2013-01-16 16:06               ` Catalin Marinas
2013-01-10  0:20 ` [PATCH 08/16] ARM: bL_platsmp.c: make sure the GIC interface of a dying CPU is disabled Nicolas Pitre
2013-01-11 18:07   ` Santosh Shilimkar
2013-01-11 19:07     ` Nicolas Pitre
2013-01-12  6:50       ` Santosh Shilimkar
2013-01-12 16:47         ` Nicolas Pitre
2013-01-13  4:37           ` Santosh Shilimkar
2013-01-14 17:53           ` Lorenzo Pieralisi
2013-01-14 16:39   ` Will Deacon
2013-01-14 16:54     ` Nicolas Pitre
2013-01-14 17:02       ` Will Deacon
2013-01-14 17:18         ` Nicolas Pitre
2013-01-14 17:24           ` Will Deacon
2013-01-14 17:56             ` Lorenzo Pieralisi
2013-01-10  0:20 ` [PATCH 09/16] ARM: vexpress: Select the correct SMP operations at run-time Nicolas Pitre
2013-01-10  0:20 ` [PATCH 10/16] ARM: vexpress: introduce DCSCB support Nicolas Pitre
2013-01-11 18:12   ` Santosh Shilimkar
2013-01-11 19:13     ` Nicolas Pitre
2013-01-12  6:52       ` Santosh Shilimkar
2013-01-10  0:20 ` [PATCH 11/16] ARM: vexpress/dcscb: add CPU use counts to the power up/down API implementation Nicolas Pitre
2013-01-10  0:20 ` [PATCH 12/16] ARM: vexpress/dcscb: do not hardcode number of CPUs per cluster Nicolas Pitre
2013-01-10  0:20 ` [PATCH 13/16] drivers: misc: add ARM CCI support Nicolas Pitre
2013-01-11 18:20   ` Santosh Shilimkar
2013-01-11 19:22     ` Nicolas Pitre
2013-01-12  6:53       ` Santosh Shilimkar
2013-01-15 18:34       ` Dave Martin
2013-01-10  0:20 ` [PATCH 14/16] ARM: TC2: ensure powerdown-time data is flushed from cache Nicolas Pitre
2013-01-10 18:50   ` Dave Martin
2013-01-10 19:13     ` Nicolas Pitre
2013-01-11 11:38       ` Dave Martin
2013-01-10  0:20 ` [PATCH 15/16] ARM: vexpress/dcscb: handle platform coherency exit/setup and CCI Nicolas Pitre
2013-01-10 12:05   ` Dave Martin
2013-01-11 18:27   ` Santosh Shilimkar
2013-01-11 19:28     ` Nicolas Pitre
2013-01-12  7:21       ` Santosh Shilimkar
2013-01-14 12:25         ` Lorenzo Pieralisi
2013-01-15  6:23           ` Santosh Shilimkar
2013-01-15 18:20             ` Dave Martin
2013-01-16  6:33               ` Santosh Shilimkar
2013-01-16 10:03                 ` Lorenzo Pieralisi
2013-01-16 10:12                   ` Santosh Shilimkar
2013-01-10  0:20 ` [PATCH 16/16] ARM: vexpress/dcscb: probe via device tree Nicolas Pitre
2013-01-10  0:46 ` [PATCH 00/16] big.LITTLE low-level CPU and cluster power management Rob Herring
2013-01-10  5:04   ` Nicolas Pitre
2013-01-10 23:01 ` Will Deacon
2013-01-14  9:56 ` Joseph Lo
2013-01-14 14:05   ` Nicolas Pitre
2013-01-15  2:44     ` Joseph Lo
2013-01-15 16:44       ` Nicolas Pitre
2013-01-16 16:02         ` Catalin Marinas
2013-01-16 21:18           ` Nicolas Pitre
2013-01-17 17:55             ` Catalin Marinas
2013-01-15 18:31     ` Dave Martin
2013-03-07  8:27 ` Pavel Machek
2013-03-07  9:12   ` Nicolas Pitre
2013-03-07  9:40     ` Pavel Machek
2013-03-07  9:56       ` Nicolas Pitre
2013-03-07 14:51         ` Pavel Machek
2013-03-07 15:42           ` Nicolas Pitre

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20130114213450.GA99454@MacBook-Pro.local \
    --to=catalin.marinas@arm.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).