Message-ID: <20130718164215.GC15992@pd.tnic>
Date:	Thu, 18 Jul 2013 18:42:15 +0200
From:	Borislav Petkov <bp@...en8.de>
To:	"Luck, Tony" <tony.luck@...el.com>
Cc:	Mauro Carvalho Chehab <mchehab@...radead.org>,
	Markus Trippelsdorf <markus@...ppelsdorf.de>,
	Ming Lei <tom.leiming@...il.com>, Linda Walsh <lkml@...nx.org>,
	Linux-Kernel <linux-kernel@...r.kernel.org>,
	Doug Thompson <dougthompson@...ssion.com>,
	"linux-edac@...r.kernel.org" <linux-edac@...r.kernel.org>
Subject: Re: BUG: key ffff880c1148c478 not in .data! (V3.10.0)

On Fri, Jul 12, 2013 at 04:19:24PM +0000, Luck, Tony wrote:
> > What would be a reasonable maximum limit for the number of memory
> > controllers, on a -EX machine?
>
> Westmere-EX has one memory controller per socket ... and there are
> glueless systems up to 8 sockets. So 8 there. Not sure if any OEM is
> building larger machines with a node controller (SGI? Not sure if they
> build their behemoths from -EP or -EX parts).
>
> Ivy Bridge ups the ante with two memory controllers on a socket. So
> plan on doubling soon.

Let's give it a second try, 16 memory controllers max:

---
>From 18fec2fd4279640b9f471c28aa3a5dc8be104273 Mon Sep 17 00:00:00 2001
From: Borislav Petkov <bp@...e.de>
Date: Fri, 12 Jul 2013 10:53:38 +0200
Subject: [PATCH] EDAC: Fix lockdep splat

Fix the following:

BUG: key ffff88043bdd0330 not in .data!
------------[ cut here ]------------
WARNING: at kernel/lockdep.c:2987 lockdep_init_map+0x565/0x5a0()
DEBUG_LOCKS_WARN_ON(1)
Modules linked in: glue_helper sb_edac(+) edac_core snd acpi_cpufreq lrw gf128mul ablk_helper iTCO_wdt evdev i2c_i801 dcdbas button cryptd pcspkr iTCO_vendor_support usb_common lpc_ich mfd_core soundcore mperf processor microcode
CPU: 2 PID: 599 Comm: modprobe Not tainted 3.10.0 #1
Hardware name: Dell Inc. Precision T3600/0PTTT9, BIOS A08 01/24/2013
 0000000000000009 ffff880439a1d920 ffffffff8160a9a9 ffff880439a1d958
 ffffffff8103d9e0 ffff88043af4a510 ffffffff81a16e11 0000000000000000
 ffff88043bdd0330 0000000000000000 ffff880439a1d9b8 ffffffff8103dacc
Call Trace:
  dump_stack
  warn_slowpath_common
  warn_slowpath_fmt
  lockdep_init_map
  ? trace_hardirqs_on_caller
  ? trace_hardirqs_on
  debug_mutex_init
  __mutex_init
  bus_register
  edac_create_sysfs_mci_device
  edac_mc_add_mc
  sbridge_probe
  pci_device_probe
  driver_probe_device
  __driver_attach
  ? driver_probe_device
  bus_for_each_dev
  driver_attach
  bus_add_driver
  driver_register
  __pci_register_driver
  ? 0xffffffffa0010fff
  sbridge_init
  ? 0xffffffffa0010fff
  do_one_initcall
  load_module
  ? unset_module_init_ro_nx
  SyS_init_module
  tracesys
---[ end trace d24a70b0d3ddf733 ]---
EDAC MC0: Giving out device to 'sbridge_edac.c' 'Sandy Bridge Socket#0': DEV 0000:3f:0e.0
EDAC sbridge: Driver loaded.

What happens is that bus_register needs a statically allocated lock_key
because it is handed in to lockdep. However, struct mem_ctl_info embeds
struct bus_type (the whole struct, not a pointer to it) which gets
dynamically allocated.

Fix this by using a statically allocated struct bus_type for the MC bus.

Cc: Mauro Carvalho Chehab <mchehab@...radead.org>
Cc: Markus Trippelsdorf <markus@...ppelsdorf.de>
Signed-off-by: Borislav Petkov <bp@...e.de>
---
 drivers/edac/edac_mc.c       |  6 ++++++
 drivers/edac/edac_mc_sysfs.c | 28 +++++++++++++++-------------
 drivers/edac/i5100_edac.c    |  2 +-
 include/linux/edac.h         |  7 ++++++-
 4 files changed, 28 insertions(+), 15 deletions(-)

diff --git a/drivers/edac/edac_mc.c b/drivers/edac/edac_mc.c
index 27e86d938262..429e971e02d7 100644
--- a/drivers/edac/edac_mc.c
+++ b/drivers/edac/edac_mc.c
@@ -48,6 +48,8 @@ static LIST_HEAD(mc_devices);
  */
 static void const *edac_mc_owner;
 
+static struct bus_type mc_bus[EDAC_MAX_MCS];
+
 unsigned edac_dimm_info_location(struct dimm_info *dimm, char *buf,
 			         unsigned len)
 {
@@ -762,6 +764,10 @@ int edac_mc_add_mc(struct mem_ctl_info *mci)
 	/* set load time so that error rate can be tracked */
 	mci->start_time = jiffies;
 
+	BUG_ON(mci->mc_idx >= EDAC_MAX_MCS);
+
+	mci->bus = &mc_bus[mci->mc_idx];
+
 	if (edac_create_sysfs_mci_device(mci)) {
 		edac_mc_printk(mci, KERN_WARNING,
 			"failed to create sysfs device\n");
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c
index ef15a7e613bc..e7c32c4f7837 100644
--- a/drivers/edac/edac_mc_sysfs.c
+++ b/drivers/edac/edac_mc_sysfs.c
@@ -370,7 +370,7 @@ static int edac_create_csrow_object(struct mem_ctl_info *mci,
 		return -ENODEV;
 
 	csrow->dev.type = &csrow_attr_type;
-	csrow->dev.bus = &mci->bus;
+	csrow->dev.bus = mci->bus;
 	device_initialize(&csrow->dev);
 	csrow->dev.parent = &mci->dev;
 	csrow->mci = mci;
@@ -605,7 +605,7 @@ static int edac_create_dimm_object(struct mem_ctl_info *mci,
 	dimm->mci = mci;
 
 	dimm->dev.type = &dimm_attr_type;
-	dimm->dev.bus = &mci->bus;
+	dimm->dev.bus = mci->bus;
 	device_initialize(&dimm->dev);
 
 	dimm->dev.parent = &mci->dev;
@@ -975,11 +975,13 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
 	 * The memory controller needs its own bus, in order to avoid
 	 * namespace conflicts at /sys/bus/edac.
 	 */
-	mci->bus.name = kasprintf(GFP_KERNEL, "mc%d", mci->mc_idx);
-	if (!mci->bus.name)
+	mci->bus->name = kasprintf(GFP_KERNEL, "mc%d", mci->mc_idx);
+	if (!mci->bus->name)
 		return -ENOMEM;
-	edac_dbg(0, "creating bus %s\n", mci->bus.name);
-	err = bus_register(&mci->bus);
+
+	edac_dbg(0, "creating bus %s\n", mci->bus->name);
+
+	err = bus_register(mci->bus);
 	if (err < 0)
 		return err;
 
@@ -988,7 +990,7 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
 	device_initialize(&mci->dev);
 
 	mci->dev.parent = mci_pdev;
-	mci->dev.bus = &mci->bus;
+	mci->dev.bus = mci->bus;
 	dev_set_name(&mci->dev, "mc%d", mci->mc_idx);
 	dev_set_drvdata(&mci->dev, mci);
 	pm_runtime_forbid(&mci->dev);
@@ -997,8 +999,8 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
 	err = device_add(&mci->dev);
 	if (err < 0) {
 		edac_dbg(1, "failure: create device %s\n", dev_name(&mci->dev));
-		bus_unregister(&mci->bus);
-		kfree(mci->bus.name);
+		bus_unregister(mci->bus);
+		kfree(mci->bus->name);
 		return err;
 	}
 
@@ -1064,8 +1066,8 @@ fail:
 	}
 fail2:
 	device_unregister(&mci->dev);
-	bus_unregister(&mci->bus);
-	kfree(mci->bus.name);
+	bus_unregister(mci->bus);
+	kfree(mci->bus->name);
 	return err;
 }
 
@@ -1098,8 +1100,8 @@ void edac_unregister_sysfs(struct mem_ctl_info *mci)
 {
 	edac_dbg(1, "Unregistering device %s\n", dev_name(&mci->dev));
 	device_unregister(&mci->dev);
-	bus_unregister(&mci->bus);
-	kfree(mci->bus.name);
+	bus_unregister(mci->bus);
+	kfree(mci->bus->name);
 }
 
 static void mc_attr_release(struct device *dev)
diff --git a/drivers/edac/i5100_edac.c b/drivers/edac/i5100_edac.c
index 1b635178cc44..157b934e8ce3 100644
--- a/drivers/edac/i5100_edac.c
+++ b/drivers/edac/i5100_edac.c
@@ -974,7 +974,7 @@ static int i5100_setup_debugfs(struct mem_ctl_info *mci)
 	if (!i5100_debugfs)
 		return -ENODEV;
 
-	priv->debugfs = debugfs_create_dir(mci->bus.name, i5100_debugfs);
+	priv->debugfs = debugfs_create_dir(mci->bus->name, i5100_debugfs);
 
 	if (!priv->debugfs)
 		return -ENOMEM;
diff --git a/include/linux/edac.h b/include/linux/edac.h
index 0b763276f619..5c6d7fbaf89e 100644
--- a/include/linux/edac.h
+++ b/include/linux/edac.h
@@ -622,7 +622,7 @@ struct edac_raw_error_desc {
  */
 struct mem_ctl_info {
 	struct device			dev;
-	struct bus_type			bus;
+	struct bus_type			*bus;
 
 	struct list_head link;	/* for global list of mem_ctl_info structs */
 
@@ -742,4 +742,9 @@ struct mem_ctl_info {
 #endif
 };
 
+/*
+ * Maximum number of memory controllers in the coherent fabric.
+ */
+#define EDAC_MAX_MCS	16
+
 #endif
-- 
1.8.3

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists