I got a bug report about a distro kernel not booting on a particular
machine. It would freeze during boot:
> ...
> Could not find start_pfn for node 1
> [boot]0015 Setup Done
> Built 2 zonelists in Node order, mobility grouping on. Total pages: 123783
> Policy zone: DMA
> Kernel command line:
> [boot]0020 XICS Init
> [boot]0021 XICS Done
> PID hash table entries: 4096 (order: 12, 32768 bytes)
> clocksource: timebase mult[7d0000] shift[22] registered
> Console: colour dummy device 80x25
> console handover: boot [udbg0] -> real [hvc0]
> Dentry cache hash table entries: 1048576 (order: 7, 8388608 bytes)
> Inode-cache hash table entries: 524288 (order: 6, 4194304 bytes)
> freeing bootmem node 0
I've reproduced this on 2.6.27.7. It is caused by commit
8f64e1f2d1 ("powerpc: Reserve in bootmem
lmb reserved regions that cross NUMA nodes").
The problem is that Jon took a loop which was (in pseudocode):
for_each_node(nid)
NODE_DATA(nid) = careful_alloc(nid);
setup_bootmem(nid);
reserve_node_bootmem(nid);
and broke it up into:
for_each_node(nid)
NODE_DATA(nid) = careful_alloc(nid);
setup_bootmem(nid);
for_each_node(nid)
reserve_node_bootmem(nid);
The issue comes in when the 'careful_alloc()' is called on a node with
no memory. It falls back to using bootmem from a previously-initialized
node. But, bootmem has not yet been reserved when Jon's patch is
applied. It gives back bogus memory (0xc000000000000000) and pukes
later in boot.
The following patch collapses the loop back together. It also breaks
the mark_reserved_regions_for_nid() code out into a function and adds
some comments. I think a huge part of introducing this bug is because
for loop was too long and hard to read.
The actual bug fix here is the:
+ if (end_pfn <= node->node_start_pfn ||
+ start_pfn >= node_end_pfn)
+ continue;
Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
numa_enforce_memory_limit tried to be smart and only call lmb_end_of_DRAM
when a memory limit was set via mem= on the command line. However,
the early boot code will also limit memory added to the lmb system
when iommu=off is specified. When this happens, the page allocator
is given pages not in the linear mapping and this results in a fatal
data reference to the unmapped page.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Adjust amount to reserve based on previous nodes for reserves spanning
multiple nodes. Check if the node active range is empty before attempting
to pass the reserve to bootmem. In practice the range shouldn't be empty,
but to be sure we check.
Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
If there are multiple reserved memory blocks via lmb_reserve() that are
contiguous addresses and on different NUMA nodes we are losing track of which
address ranges to reserve in bootmem on which node. I discovered this
when I recently got to try 16GB huge pages on a system with more then 2 nodes.
When scanning the device tree in early boot we call lmb_reserve() with
the addresses of the 16G pages that we find so that the memory doesn't
get used for something else. For example the addresses for the pages
could be 4000000000, 4400000000, 4800000000, 4C00000000, etc - 8 pages,
one on each of eight nodes. In the lmb after all the pages have been
reserved it will look something like the following:
lmb_dump_all:
memory.cnt = 0x2
memory.size = 0x3e80000000
memory.region[0x0].base = 0x0
.size = 0x1e80000000
memory.region[0x1].base = 0x4000000000
.size = 0x2000000000
reserved.cnt = 0x5
reserved.size = 0x3e80000000
reserved.region[0x0].base = 0x0
.size = 0x7b5000
reserved.region[0x1].base = 0x2a00000
.size = 0x78c000
reserved.region[0x2].base = 0x328c000
.size = 0x43000
reserved.region[0x3].base = 0xf4e8000
.size = 0xb18000
reserved.region[0x4].base = 0x4000000000
.size = 0x2000000000
The reserved.region[0x4] contains the 16G pages. In
arch/powerpc/mm/num.c: do_init_bootmem() we loop through each of the
node numbers looking for the reserved regions that belong to the
particular node. It is not able to identify region 0x4 as being a part
of each of the 8 nodes. It is assuming that a reserved region is only
on a single node.
This patch takes out the reserved region loop from inside
the loop that goes over each node. It looks up the active region containing
the start of the reserved region. If it extends past that active region then
it adjusts the size and gets the next active region containing it.
Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
There are a lot of places that define either a single bootmem descriptor or an
array of them. Use only one central array with MAX_NUMNODES items instead.
Signed-off-by: Johannes Weiner <hannes@saeurebad.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Kyle McMartin <kyle@parisc-linux.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Update the association of a memory section with a numa node that
occurs during hotplug add of a memory section. This adds a check in
the hot_add_scn_to_nid() routine for the
ibm,dynamic-reconfiguration-memory node in the device tree. If
present the new hot_add_drconf_scn_to_nid() routine is invoked, which
can properly parse the ibm,dynamic-reconfiguration-memory node of the
device tree and make the proper numa node associations.
This also introduces the valid_hot_add_scn() routine as a helper
function for code that is common to the hot_add_scn_to_nid() and
hot_add_drconf_scn_to_nid() routines.
Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This splits off several pieces of code that parse the
ibm,dynamic-reconfiguration-memory node of the device tree into separate
helper routines. This is in preparation for the next commit that will
use these helper routines. There are no functional changes in this patch.
Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
numa.c requires routines declared in linux/of.h, so should include it.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patchset adds a flags variable to reserve_bootmem() and uses the
BOOTMEM_EXCLUSIVE flag in crashkernel reservation code to detect collisions
between crashkernel area and already used memory.
This patch:
Change the reserve_bootmem() function to accept a new flag BOOTMEM_EXCLUSIVE.
If that flag is set, the function returns with -EBUSY if the memory already
has been reserved in the past. This is to avoid conflicts.
Because that code runs before SMP initialisation, there's no race condition
inside reserve_bootmem_core().
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix powerpc build]
Signed-off-by: Bernhard Walle <bwalle@suse.de>
Cc: <linux-arch@vger.kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Here's a dumb simple implementation of fake NUMA nodes for PowerPC.
Fake NUMA nodes can be specified using the following command line
option
numa=fake=<node range>
node range is of the format <range1>,<range2>,...<rangeN>
Each of the rangeX parameters is passed using memparse(). I find the
patch useful for fake NUMA emulation on my simple PowerPC machine.
I've tested it on a numa box with the following arguments
numa=fake=512M
numa=fake=512M,768M
numa=fake=256M,512M mem=512M
numa=fake=1G mem=768M
numa=fake=
without any numa= argument
The other side-effect introduced by this patch is that; in the case
where we don't have NUMA information, we now set a node online after
adding each LMB. This node could very well be node 0, but in the case
that we enable fake NUMA nodes, when we cross node boundaries, we need
to set the new node online.
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Some new machines use the "ibm,dynamic-reconfiguration-memory" property
to provide memory layout information, rather than via memory nodes.
There is a bug in the code to parse this property for start addresses
over 4GB; we store the start address in an unsigned int, which means
we throw away the high bits and add apparently duplicate regions.
This results in a BUG() in free_bootmem_core(). This fixes it by
using an unsigned long instead.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Since nonboot CPUs are now disabled after tasks and devices have been
frozen and the CPU hotplug infrastructure is used for this purpose, we need
special CPU hotplug notifications that will help the CPU-hotplug-aware
subsystems distinguish normal CPU hotplug events from CPU hotplug events
related to a system-wide suspend or resume operation in progress. This
patch introduces such notifications and causes them to be used during
suspend and resume transitions. It also changes all of the
CPU-hotplug-aware subsystems to take these notifications into consideration
(for now they are handled in the same way as the corresponding "normal"
ones).
[oleg@tv-sign.ru: cleanups]
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Gautham R Shenoy <ego@in.ibm.com>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This is more consistent and gets us closer to the Sparc code.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This is more consistent and gets us closer to the Sparc code.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Arch-independent zone-sizing is using indices instead of symbolic names to
offset within an array related to zones (max_zone_pfns). The unintended
impact is that ZONE_DMA and ZONE_NORMAL is initialised on powerpc instead
of ZONE_DMA and ZONE_HIGHMEM when CONFIG_HIGHMEM is set. As a result, the
the machine fails to boot but will boot with CONFIG_HIGHMEM turned off.
The following patch properly initialises the max_zone_pfns[] array and uses
symbolic names instead of indices in each architecture using
arch-independent zone-sizing. Two users have successfully booted their
powerpcs with it (one an ibook G4). It has also been boot tested on x86,
x86_64, ppc64 and ia64. Please merge for 2.6.19-rc2.
Credit to Benjamin Herrenschmidt for identifying the bug and rolling the
first fix. Additional credit to Johannes Berg and Andreas Schwab for
reporting the problem and testing on powerpc.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Now that get_property() returns a void *, there's no need to cast its
return value. Also, treat the return value as const, so we can
constify get_property later.
powerpc core changes.
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Make notifier_blocks associated with cpu_notifier as __cpuinitdata.
__cpuinitdata makes sure that the data is init time only unless
CONFIG_HOTPLUG_CPU is defined.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Change of_node_to_nid() to traverse the device tree, looking for a numa id.
Cell uses this to assign ids to SPUs, which are children of the CPU node.
Existing users of of_node_to_nid() are altered to use of_node_to_nid_single(),
which doesn't do the traversal.
Export an attach_sysdev_to_node() function, allowing system devices (eg.
SPUs) to link themselves into the numa topology in sysfs.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Quieten some of the debug ram config output. we already print out available
memory at KERN_INFO level.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The return statement is to prevent `warning: 'nid' might be used uninitialized
in this function'.
Cc: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
We can plug the boot cpu into its node independently of whether numa
topology is detected. And numa_setup_cpu does the right thing for all
cases now, so remove special-casing for non-numa from the cpu hotplug
callback.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The powerpc numa code unconditionally onlines all nodes from 0 to the
highest node id found, regardless of whether cpus or memory are
present in the nodes. This wastes 8K per node and complicates some
cpu and memory hotplug situations, such as adding a resource that
doesn't map to one of the nodes discovered at boot.
Set nodes online as resources are scanned. Fall back to node 0 only
when we're sure this isn't a NUMA machine.
Instead of defaulting to node 0 for cases of hot-adding a resource
which doesn't belong to any initialized node, assign it to the first
online node.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Code to handle Power4's invalid node id (0xffff) is duplicated for cpu
and memory. Better to handle this case in one place --
of_node_to_nid. Overall behavior should be unchanged.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Since we effectively treat the domain ids given to us by firmare as
logical node ids, make this explicit (basically s/numa_domain/nid/).
No functional changes, only variable and function names are modified.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
map_cpu_to_node does not need to be inline, it is never called in a
hot path.
map_cpu_to_node, numa_setup_cpu, and find_cpu_node can be marked
__cpuinit, as they are never used after boot if CONFIG_HOTPLUG_CPU=n.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Add debug statement for map_cpu_to_node; it's useful for cpu hotplug.
Clarify debug statement about not finding the numa reference points
property.
Don't print a meaningless associativity depth (-1) on non-numa systems.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
At boot, the numa code is assigning boot_cpuid to node 0
unconditionally. Basically, numa_setup_cpu is being stupid about it,
but this is the minimal fix -- just call numa_setup_cpu(boot_cpuid)
later, after all nodes have been set online.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Currently most callers of lmb_alloc() don't check if it worked or not, if it
ever does weird bad things will probably happen. The few callers who do check
just panic or BUG_ON.
So make lmb_alloc() panic internally, to catch bugs at the source. The few
callers who did check the result no longer need to.
The only caller that did anything interesting with the return result was
careful_allocation(). For it we create __lmb_alloc_base() which _doesn't_ panic
automatically, a little messy, but passable.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The system will oops if an attempt is made to add memory to an
empty node/zone. This patch prevents adding memory to an empty
node. The code to dynamically add a node/zone is non-trivial.
This patch is temporary and will be removed when the ability
to dynamically add a node/zone is complete.
Signed-off-by: Mike Kravetz <kravetz@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The pre-parsed addrs/n_addrs fields in struct device_node are finally
gone. Remove the dodgy heuristics that did that parsing at boot and
remove the fields themselves since we now have a good replacement with
the new OF parsing code. This patch also fixes a bunch of drivers to use
the new code instead, so that at least pmac32, pseries, iseries and g5
defconfigs build.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
We used to print a NUMA cpu summary at boot before the hotplug cpu code
was added. This has been useful for catching machine configuration as
well as firmware bugs in the past.
This patch restores that functionality. An example of the output is:
Node 0 CPUs: 0-7
Node 1 CPUs: 8-15
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Milton has proposed that we should support a "linux,usable-memory" property
on memory nodes which describes, in preference to "reg", the regions of memory
Linux should use.
This facility is required for kdump to inform the second kernel which memory
it should use.
Signed-off-by: Haren Myneni <haren@us.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This places dynamically added memory within the appropriate
numa node. A new routine hot_add_scn_to_nid() replicates most of
the memory scanning code in parse_numa_properties().
Signed-off-by: Mike Kravetz <kravetz@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Here is an updated version of the patch that panics if no memory is
found as Nathan suggested. I'm still concerned that panic strings
(not just the one added here) at this stage of booting do not show
up on my system. But, that is an issue separate from this patch.
Combine get_mem_*_cells() routines to avoid multiple memory node
lookups. Added missing of_node_put() call. Changed variable names
to help with some confusion as to meaning.
Signed-off-by: Mike Kravetz <kravetz@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
I started to add missing of_node_put() calls to the routines that
determine the number of cells for memory. Decided to combine the
routines instead of making separate node lookups. Changed variable
names to help with some confusion as to meaning.
Signed-off-by: Mike Kravetz <kravetz@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This bug exists in the current code and prevents machines from booting
with numa enabled if there is a node that does not contain memory.
Workaround is to boot with 'numa=off'. Looks like a simple typo.
Signed-off-by: Mike Kravetz <kravetz@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch should fix the crashes we have been seeing on 64-bit
powerpc systems with a memory hole when sparsemem is enabled.
I'd appreciate it if people who know more about NUMA and sparsemem
than me could look over it.
There were two bugs. The first was that if NUMA was enabled but there
was no NUMA information for the machine, the setup_nonnuma() function
was adding a single region, assuming memory was contiguous. The
second was that the loops in mem_init() and show_mem() assumed that
all pages within the span of a pgdat were valid (had a valid struct
page).
I also fixed the incorrect setting of num_physpages that Mike Kravetz
pointed out.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Convert to sparsemem and remove all the discontigmem code in the
process. This has a few advantages:
- The old numa_memory_lookup_table can go away
- All the arch specific discontigmem magic can go away
We also remove the triple pass of memory properties and instead create a
list of per node extents that we iterate through. A final cleanup would
be to change our lmb code to store extents per node, then we can reuse
that information in the numa code.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Remove ppc64 specific version of nr_cpus_node and use the generic one
provided.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Mostly this involves adding #include <asm/smp.h>, since that defines
things like boot_cpuid[_phys] and [gs]et_hard_smp_processor_id, which
are SMP-related but still needed on UP. This incorporates fixes
posted by Olof Johansson and Heikki Lindholm.
Signed-off-by: Paul Mackerras <paulus@samba.org>
We had a static memory_limit in prom.c, and then another one defined
in setup_64.c and used in numa.c, which resulted in the kernel crashing
when mem=xxx was given on the command line. This puts the declaration
in system.h and the definition in mem.c. This also moves the
definition of tce_alloc_start/end out of setup_64.c.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Dave Hansen