Add support for using multiple hugepage sizes simultaneously
on mainline. Currently, support for 256M has been added which
can be used along with 8M pages.
Page tables are set like this (e.g. for 256M page):
VA + (8M * x) -> PA + (8M * x) (sz bit = 256M) where x in [0, 31]
and TSB is set similarly:
VA + (4M * x) -> PA + (4M * x) (sz bit = 256M) where x in [0, 63]
- Testing
Tested on Sonoma (which supports 256M pages) by running stream
benchmark instances in parallel: one instance uses 8M pages and
another uses 256M pages, consuming 48G each.
Boot params used:
default_hugepagesz=256M hugepagesz=256M hugepages=300 hugepagesz=8M
hugepages=10000
Signed-off-by: Nitin Gupta <nitin.m.gupta@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This was entirely automated, using the script by Al:
PATT='^[[:blank:]]*#[[:blank:]]*include[[:blank:]]*<asm/uaccess.h>'
sed -i -e "s!$PATT!#include <linux/uaccess.h>!" \
$(git grep -l "$PATT"|grep -v ^include/linux/uaccess.h)
to do the replacement at the end of the merge window.
Requested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A compile warning is introduced by a commit to fix the find_node().
This patch fix the compile warning by moving find_node() into __init
section. Because find_node() is only used by memblock_nid_range() which
is only used by a __init add_node_ranges(). find_node() and
memblock_nid_range() should also be inside __init section.
Signed-off-by: Thomas Tai <thomas.tai@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When booting up LDOM, find_node() warns that a physical address
doesn't match a NUMA node.
WARNING: CPU: 0 PID: 0 at arch/sparc/mm/init_64.c:835
find_node+0xf4/0x120 find_node: A physical address doesn't
match a NUMA node rule. Some physical memory will be
owned by node 0.Modules linked in:
CPU: 0 PID: 0 Comm: swapper Not tainted 4.9.0-rc3 #4
Call Trace:
[0000000000468ba0] __warn+0xc0/0xe0
[0000000000468c74] warn_slowpath_fmt+0x34/0x60
[00000000004592f4] find_node+0xf4/0x120
[0000000000dd0774] add_node_ranges+0x38/0xe4
[0000000000dd0b1c] numa_parse_mdesc+0x268/0x2e4
[0000000000dd0e9c] bootmem_init+0xb8/0x160
[0000000000dd174c] paging_init+0x808/0x8fc
[0000000000dcb0d0] setup_arch+0x2c8/0x2f0
[0000000000dc68a0] start_kernel+0x48/0x424
[0000000000dcb374] start_early_boot+0x27c/0x28c
[0000000000a32c08] tlb_fixup_done+0x4c/0x64
[0000000000027f08] 0x27f08
It is because linux use an internal structure node_masks[] to
keep the best memory latency node only. However, LDOM mdesc can
contain single latency-group with multiple memory latency nodes.
If the address doesn't match the best latency node within
node_masks[], it should check for an alternative via mdesc.
The warning message should only be printed if the address
doesn't match any node_masks[] nor within mdesc. To minimize
the impact of searching mdesc every time, the last matched
mask and index is stored in a variable.
Signed-off-by: Thomas Tai <thomas.tai@oracle.com>
Reviewed-by: Chris Hyser <chris.hyser@oracle.com>
Reviewed-by: Liam Merwick <liam.merwick@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These files were only including module.h for exception table
related functions. We've now separated that content out into its
own file "extable.h" so now move over to that and avoid all the
extra header content in module.h that we don't really need to compile
these files.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: sparclinux@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, irq stack bootmem is allocated for all possible cpus
before nr_cpus value changes the list of possible cpus. As a result,
there is unnecessary wastage of bootmemory.
Move the irq stack bootmem allocation so that it happens after
possible cpu list is modified based on nr_cpus value.
Signed-off-by: Atish Patra <atish.patra@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Vijay Kumar <vijay.ac.kumar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To fix:
WARNING: vmlinux.o(.text.unlikely+0x580): Section mismatch in
reference from the function find_numa_latencies_for_group() to the
function .init.text:find_mlgroup()
The function find_numa_latencies_for_group() references the
function __init find_mlgroup(). This is often because
find_numa_latencies_for_group lacks a __init annotation or the
annotation of find_mlgroup is wrong.
It turns out find_numa_latencies_for_group is only called from:
static int __init numa_parse_mdesc(void)
and hence we can tag find_numa_latencies_for_group with __init.
In doing so we see that find_best_numa_node_for_mlgroup is only
called from within __init and hence can also be marked with __init.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Chris Hyser <chris.hyser@oracle.com>
Cc: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Cc: sparclinux@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For PMD aligned (8M) hugepages, we currently allocate
all four page table levels which is wasteful. We now
allocate till PMD level only which saves memory usage
from page tables.
Also, when freeing page table for 8M hugepage backed region,
make sure we don't try to access non-existent PTE level.
Orabug: 22630259
Signed-off-by: Nitin Gupta <nitin.m.gupta@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
do_sparc64_fault() calculates both the base and huge page RSS sizes and
uses this information in calls to tsb_grow(). The calculation for base
page TSB size is not correct if the task uses hugetlb pages. hugetlb
pages are not accounted for in RSS, therefore the call to get_mm_rss(mm)
does not include hugetlb pages. However, the number of pages based on
huge_pte_count (which does include hugetlb pages) is subtracted from
this value. This will result in an artificially small and often negative
RSS calculation. The base TSB size is then often set to max_tsb_size
as the passed RSS is unsigned, so a negative value looks really big.
THP pages are also accounted for in huge_pte_count, and THP pages are
accounted for in RSS so the calculation in do_sparc64_fault() is correct
if a task only uses THP pages.
A single huge_pte_count is not sufficient for TSB sizing if both hugetlb
and THP pages can be used. Instead of a single counter, use two: one
for hugetlb and one for THP.
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the third version of the patchset previously sent [1]. I have
basically only rebased it on top of 4.7-rc1 tree and dropped "dm: get
rid of superfluous gfp flags" which went through dm tree. I am sending
it now because it is tree wide and chances for conflicts are reduced
considerably when we want to target rc2. I plan to send the next step
and rename the flag and move to a better semantic later during this
release cycle so we will have a new semantic ready for 4.8 merge window
hopefully.
Motivation:
While working on something unrelated I've checked the current usage of
__GFP_REPEAT in the tree. It seems that a majority of the usage is and
always has been bogus because __GFP_REPEAT has always been about costly
high order allocations while we are using it for order-0 or very small
orders very often. It seems that a big pile of them is just a
copy&paste when a code has been adopted from one arch to another.
I think it makes some sense to get rid of them because they are just
making the semantic more unclear. Please note that GFP_REPEAT is
documented as
* __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
* _might_ fail. This depends upon the particular VM implementation.
while !costly requests have basically nofail semantic. So one could
reasonably expect that order-0 request with __GFP_REPEAT will not loop
for ever. This is not implemented right now though.
I would like to move on with __GFP_REPEAT and define a better semantic
for it.
$ git grep __GFP_REPEAT origin/master | wc -l
111
$ git grep __GFP_REPEAT | wc -l
36
So we are down to the third after this patch series. The remaining
places really seem to be relying on __GFP_REPEAT due to large allocation
requests. This still needs some double checking which I will do later
after all the simple ones are sorted out.
I am touching a lot of arch specific code here and I hope I got it right
but as a matter of fact I even didn't compile test for some archs as I
do not have cross compiler for them. Patches should be quite trivial to
review for stupid compile mistakes though. The tricky parts are usually
hidden by macro definitions and thats where I would appreciate help from
arch maintainers.
[1] http://lkml.kernel.org/r/1461849846-27209-1-git-send-email-mhocko@kernel.org
This patch (of 19):
__GFP_REPEAT has a rather weak semantic but since it has been introduced
around 2.6.12 it has been ignored for low order allocations. Yet we
have the full kernel tree with its usage for apparently order-0
allocations. This is really confusing because __GFP_REPEAT is
explicitly documented to allow allocation failures which is a weaker
semantic than the current order-0 has (basically nofail).
Let's simply drop __GFP_REPEAT from those places. This would allow to
identify place which really need allocator to retry harder and formulate
a more specific semantic for what the flag is supposed to do actually.
Link: http://lkml.kernel.org/r/1464599699-30131-2-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Chen Liqin <liqin.linux@gmail.com>
Cc: Chris Metcalf <cmetcalf@mellanox.com> [for tile]
Cc: Guan Xuetao <gxt@mprc.pku.edu.cn>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: John Crispin <blogic@openwrt.org>
Cc: Lennox Wu <lennox.wu@gmail.com>
Cc: Ley Foon Tan <lftan@altera.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
On cheetahplus chips we take the ctx_alloc_lock in order to
modify the TLB lookup parameters for the indexed TLBs, which
are stored in the context register.
This is called with interrupts disabled, however ctx_alloc_lock
is an IRQ safe lock, therefore we must take acquire/release it
properly with spin_{lock,unlock}_irq().
Reported-by: Meelis Roos <mroos@linux.ee>
Tested-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: David S. Miller <davem@davemloft.net>
During hugepage map/unmap, TSB and TLB flushes are currently
issued at every PAGE_SIZE'd boundary which is unnecessary.
We now issue the flush at REAL_HPAGE_SIZE boundaries only.
Without this patch workloads which unmap a large hugepage
backed VMA region get CPU lockups due to excessive TLB
flush calls.
Orabug: 22365539, 22643230, 22995196
Signed-off-by: Nitin Gupta <nitin.m.gupta@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add code to recognize SPARC-Sonoma cpu correctly and update cpu hardware
caps and cpu distribution map. SPARC-Sonoma is based upon SPARC-M7 core
along with additional PCI functions added on and is reported by firmware
as "SPARC-SN".
Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
Acked-by: Allen Pais <allen.pais@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set IORESOURCE_SYSTEM_RAM in flags of resource ranges with
"System RAM", "Kernel code", "Kernel data", and "Kernel bss".
Note that:
- IORESOURCE_SYSRAM (i.e. modifier bit) is set in flags when
IORESOURCE_MEM is already set. IORESOURCE_SYSTEM_RAM is defined
as (IORESOURCE_MEM|IORESOURCE_SYSRAM).
- Some archs do not set 'flags' for children nodes, such as
"Kernel code". This patch does not change 'flags' in this
case.
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luis R. Rodriguez <mcgrof@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: linux-arch@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-mips@linux-mips.org
Cc: linux-mm <linux-mm@kvack.org>
Cc: linux-parisc@vger.kernel.org
Cc: linux-s390@vger.kernel.org
Cc: linux-sh@vger.kernel.org
Cc: linuxppc-dev@lists.ozlabs.org
Cc: sparclinux@vger.kernel.org
Link: http://lkml.kernel.org/r/1453841853-11383-7-git-send-email-bp@alien8.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Orabug: 22495713
Currently, NUMA node distance matrix is initialized only
when a machine descriptor (MD) exists. However, sun4u
machines (e.g. Sun Blade 2500) do not have an MD and thus
distance values were left uninitialized. The initialization
is now moved such that it happens on both sun4u and sun4v.
Signed-off-by: Nitin Gupta <nitin.m.gupta@oracle.com>
Tested-by: Mikael Pettersson <mikpelinux@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some high end Intel Xeon systems report uncorrectable memory errors as a
recoverable machine check. Linux has included code for some time to
process these and just signal the affected processes (or even recover
completely if the error was in a read only page that can be replaced by
reading from disk).
But we have no recovery path for errors encountered during kernel code
execution. Except for some very specific cases were are unlikely to ever
be able to recover.
Enter memory mirroring. Actually 3rd generation of memory mirroing.
Gen1: All memory is mirrored
Pro: No s/w enabling - h/w just gets good data from other side of the
mirror
Con: Halves effective memory capacity available to OS/applications
Gen2: Partial memory mirror - just mirror memory begind some memory controllers
Pro: Keep more of the capacity
Con: Nightmare to enable. Have to choose between allocating from
mirrored memory for safety vs. NUMA local memory for performance
Gen3: Address range partial memory mirror - some mirror on each memory
controller
Pro: Can tune the amount of mirror and keep NUMA performance
Con: I have to write memory management code to implement
The current plan is just to use mirrored memory for kernel allocations.
This has been broken into two phases:
1) This patch series - find the mirrored memory, use it for boot time
allocations
2) Wade into mm/page_alloc.c and define a ZONE_MIRROR to pick up the
unused mirrored memory from mm/memblock.c and only give it out to
select kernel allocations (this is still being scoped because
page_alloc.c is scary).
This patch (of 3):
Add extra "flags" to memblock to allow selection of memory based on
attribute. No functional changes
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Xiexiuqi <xiexiuqi@huawei.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
sparc: Resolve conflict between sparc v9 and M7 on usage of bit 9 of TTE
Bit 9 of TTE is CV (Cacheable in V-cache) on sparc v9 processor while
the same bit 9 is MCDE (Memory Corruption Detection Enable) on M7
processor. This creates a conflicting usage of the same bit. Kernel
sets TTE.cv bit on all pages for sun4v architecture which works well
for sparc v9 but enables memory corruption detection on M7 processor
which is not the intent. This patch adds code to determine if kernel
is running on M7 processor and takes steps to not enable memory
corruption detection in TTE erroneously.
Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce faulthandler_disabled() and use it to check for irq context and
disabled pagefaults (via pagefault_disable()) in the pagefault handlers.
Please note that we keep the in_atomic() checks in place - to detect
whether in irq context (in which case preemption is always properly
disabled).
In contrast, preempt_disable() should never be used to disable pagefaults.
With !CONFIG_PREEMPT_COUNT, preempt_disable() doesn't modify the preempt
counter, and therefore the result of in_atomic() differs.
We validate that condition by using might_fault() checks when calling
might_sleep().
Therefore, add a comment to faulthandler_disabled(), describing why this
is needed.
faulthandler_disabled() and pagefault_disable() are defined in
linux/uaccess.h, so let's properly add that include to all relevant files.
This patch is based on a patch from Thomas Gleixner.
Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David.Laight@ACULAB.COM
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bigeasy@linutronix.de
Cc: borntraeger@de.ibm.com
Cc: daniel.vetter@intel.com
Cc: heiko.carstens@de.ibm.com
Cc: herbert@gondor.apana.org.au
Cc: hocko@suse.cz
Cc: hughd@google.com
Cc: mst@redhat.com
Cc: paulus@samba.org
Cc: ralf@linux-mips.org
Cc: schwidefsky@de.ibm.com
Cc: yang.shi@windriver.com
Link: http://lkml.kernel.org/r/1431359540-32227-7-git-send-email-dahi@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
/proc/kcore investigates the "System RAM" elements in /proc/iomem to
initialize it's memory tables. Therefore we have to register them
before it tries to do so. kcore uses device_initcall() so let's
use arch_initcall() for the registry.
Also we need ARCH_PROC_KCORE_TEXT to get the virtual addresses of
the kernel image correct.
Reported-by: David Ahern <david.ahern@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now, we have prepared to avoid using debug-pagealloc in boottime. So
introduce new kernel-parameter to disable debug-pagealloc in boottime, and
makes related functions to be disabled in this case.
Only non-intuitive part is change of guard page functions. Because guard
page is effective only if debug-pagealloc is enabled, turning off
according to debug-pagealloc is reasonable thing to do.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
swapper_low_pmd_dir and swapper_pud_dir are actually completely
useless and unnecessary.
We just need swapper_pg_dir[]. Naturally the other page table chunks
will be allocated on an as-needed basis. Since the kernel actually
accesses these tables in the PAGE_OFFSET view, there is not even a TLB
locality advantage of placing them in the kernel image.
Use the hard coded vmlinux.ld.S slot for swapper_pg_dir which is
naturally page aligned.
Increase MAX_BANKS to 1024 in order to handle heavily fragmented
virtual guests.
Even with this MAX_BANKS increase, the kernel is 20K+ smaller.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
In order to accomodate embedded per-cpu allocation with large numbers
of cpus and numa nodes, we have to use as much virtual address space
as possible for the vmalloc region. Otherwise we can get things like:
PERCPU: max_distance=0x380001c10000 too large for vmalloc space 0xff00000000
So, once we select a value for PAGE_OFFSET, derive the size of the
vmalloc region based upon that.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
Make sure, at compile time, that the kernel can properly support
whatever MAX_PHYS_ADDRESS_BITS is defined to.
On M7 chips, use a max_phys_bits value of 49.
Based upon a patch by Bob Picco.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
For sparse memory configurations, the vmemmap array behaves terribly
and it takes up an inordinate amount of space in the BSS section of
the kernel image unconditionally.
Just build huge PMDs and look them up just like we do for TLB misses
in the vmalloc area.
Kernel BSS shrinks by about 2MB.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like
DEBUG_PAGEALLOC stop working because the 3-level page tables only
can cover up to 43 bits.
Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to
47, several statically allocated tables became enormous.
Compounding this is that we will need to support up to 49 bits of
physical addressing for M7 chips.
The two tables in question are sparc64_valid_addr_bitmap and
kpte_linear_bitmap.
The first holds a bitmap, with 1 bit for each 4MB chunk of physical
memory, indicating whether that chunk actually exists in the machine
and is valid.
The second table is a set of 2-bit values which tell how large of a
mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB
chunk of ram in the system.
These tables are huge and take up an enormous amount of the BSS
section of the sparc64 kernel image. Specifically, the
sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K.
So let's solve the space wastage and the DEBUG_PAGEALLOC problem
at the same time, by using the kernel page tables (as designed) to
manage this information.
We have to keep using large mappings when DEBUG_PAGEALLOC is disabled,
and we do this by encoding huge PMDs and PUDs.
On a T4-2 with 256GB of ram the kernel page table takes up 16K with
DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this
memory is dynamically allocated at run time rather than coded
statically into the kernel image.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
As currently coded the KTSB accesses in the kernel only support up to
47 bits of physical addressing.
Adjust the instruction and patching sequence in order to support
arbitrary 64 bits addresses.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
Now that we use 4-level page tables, we can provide up to 53-bits of
virtual address space to the user.
Adjust the VA hole based upon the capabilities of the cpu type probed.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
This has become necessary with chips that support more than 43-bits
of physical addressing.
Based almost entirely upon a patch by Bob Picco.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
The "mem" boot option can result in many unexpected consequences. This patch
attempts to prevent boot hangs which have been experienced on T4-4 and T5-8.
Basically the boot loader allocates vmlinuz and initrd higher in available
OBP physical memory. For example, on a 2Tb T5-8 it isn't possible to boot
with mem=20G.
The patch utilizes memblock to avoid reserved regions and trim memory which
is only free. Other improvements are possible for a multi-node machine.
This is a snippet of the boot log with mem=20G on T5-8 with the patch applied:
MEMBLOCK configuration: <- before memory reduction
memory size = 0x1ffad6ce000 reserved size = 0xa1adf44
memory.cnt = 0xb
memory[0x0] [0x00000030400000-0x00003fdde47fff], 0x3fada48000 bytes
memory[0x1] [0x00003fdde4e000-0x00003fdde4ffff], 0x2000 bytes
memory[0x2] [0x00080000000000-0x00083fffffffff], 0x4000000000 bytes
memory[0x3] [0x00100000000000-0x00103fffffffff], 0x4000000000 bytes
memory[0x4] [0x00180000000000-0x00183fffffffff], 0x4000000000 bytes
memory[0x5] [0x00200000000000-0x00203fffffffff], 0x4000000000 bytes
memory[0x6] [0x00280000000000-0x00283fffffffff], 0x4000000000 bytes
memory[0x7] [0x00300000000000-0x00303fffffffff], 0x4000000000 bytes
memory[0x8] [0x00380000000000-0x00383fffc71fff], 0x3fffc72000 bytes
memory[0x9] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes
memory[0xa] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes
reserved.cnt = 0x2
reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes
reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes
...
MEMBLOCK configuration: <- after reduction of memory
memory size = 0x50a1adf44 reserved size = 0xa1adf44
memory.cnt = 0x4
memory[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes
memory[0x1] [0x00380004000000-0x0038050d01d74a], 0x50901d74b bytes
memory[0x2] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes
memory[0x3] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes
reserved.cnt = 0x2
reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes
reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes
...
Early memory node ranges
node 7: [mem 0x380000000000-0x38000117dfff]
node 7: [mem 0x380004000000-0x380f0d01bfff]
node 7: [mem 0x383fffc92000-0x383fffca1fff]
node 7: [mem 0x383fffcb4000-0x383fffcb5fff]
Could not find start_pfn for node 0
Could not find start_pfn for node 1
Could not find start_pfn for node 2
Could not find start_pfn for node 3
Could not find start_pfn for node 4
Could not find start_pfn for node 5
Could not find start_pfn for node 6
.
The patch was tested on T4-1, T5-8 and Jalap?no.
Cc: sparclinux@vger.kernel.org
Signed-off-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We have seen an issue with guest boot into LDOM that causes early boot failures
because of no matching rules for node identitity of the memory. I analyzed this
on my T4 and concluded there might not be a solution. I saw the issue in
mainline too when booting into the control/primary domain - with guests
configured. Note, this could be a firmware bug on some older machines.
I'll provide a full explanation of the issues below. Should we not find a
matching BEST latency group for a real address (RA) then we will assume node 0.
On the T4-2 here with the information provided I can't see an alternative.
Technically the LDOM shown below should match the MBLOCK to the
favorable latency group. However other factors must be considered too. Were
the memory controllers configured "fine" grained interleave or "coarse"
grain interleaved - T4. Also should a "group" MD node be considered a NUMA
node?
There has to be at least one Machine Description (MD) "group" and hence one
NUMA node. The group can have one or more latency groups (lg) - more than one
memory controller. The current code chooses the smallest latency as the most
favorable per group. The latency and lg information is in MLGROUP below.
MBLOCK is the base and size of the RAs for the machine as fetched from OBP
/memory "available" property. My machine has one MBLOCK but more would be
possible - with holes?
For a T4-2 the following information has been gathered:
with LDOM guest
MEMBLOCK configuration:
memory size = 0x27f870000
memory.cnt = 0x3
memory[0x0] [0x00000020400000-0x0000029fc67fff], 0x27f868000 bytes
memory[0x1] [0x0000029fd8a000-0x0000029fd8bfff], 0x2000 bytes
memory[0x2] [0x0000029fd92000-0x0000029fd97fff], 0x6000 bytes
reserved.cnt = 0x2
reserved[0x0] [0x00000020800000-0x000000216c15c0], 0xec15c1 bytes
reserved[0x1] [0x00000024800000-0x0000002c180c1e], 0x7980c1f bytes
MBLOCK[0]: base[20000000] size[280000000] offset[0]
(note: "base" and "size" reported in "MBLOCK" encompass the "memory[X]" values)
(note: (RA + offset) & mask = val is the formula to detect a match for the
memory controller. should there be no match for find_node node, a return
value of -1 resulted for the node - BAD)
There is one group. It has these forward links
MLGROUP[1]: node[545] latency[1f7e8] match[200000000] mask[200000000]
MLGROUP[2]: node[54d] latency[2de60] match[0] mask[200000000]
NUMA NODE[0]: node[545] mask[200000000] val[200000000] (latency[1f7e8])
(note: "val" is the best lg's (smallest latency) "match")
no LDOM guest - bare metal
MEMBLOCK configuration:
memory size = 0xfdf2d0000
memory.cnt = 0x3
memory[0x0] [0x00000020400000-0x00000fff6adfff], 0xfdf2ae000 bytes
memory[0x1] [0x00000fff6d2000-0x00000fff6e7fff], 0x16000 bytes
memory[0x2] [0x00000fff766000-0x00000fff771fff], 0xc000 bytes
reserved.cnt = 0x2
reserved[0x0] [0x00000020800000-0x00000021a04580], 0x1204581 bytes
reserved[0x1] [0x00000024800000-0x0000002c7d29fc], 0x7fd29fd bytes
MBLOCK[0]: base[20000000] size[fe0000000] offset[0]
there are two groups
group node[16d5]
MLGROUP[0]: node[1765] latency[1f7e8] match[0] mask[200000000]
MLGROUP[3]: node[177d] latency[2de60] match[200000000] mask[200000000]
NUMA NODE[0]: node[1765] mask[200000000] val[0] (latency[1f7e8])
group node[171d]
MLGROUP[2]: node[1775] latency[2de60] match[0] mask[200000000]
MLGROUP[1]: node[176d] latency[1f7e8] match[200000000] mask[200000000]
NUMA NODE[1]: node[176d] mask[200000000] val[200000000] (latency[1f7e8])
(note: for this two "group" bare metal machine, 1/2 memory is in group one's
lg and 1/2 memory is in group two's lg).
Cc: sparclinux@vger.kernel.org
Signed-off-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based almost entirely upon a patch by Christopher Alexander Tobias
Schulze.
In commit db64fe0225 ("mm: rewrite vmap
layer") lazy VMAP tlb flushing was added to the vmalloc layer. This
causes problems on sparc64.
Sparc64 has two VMAP mapped regions and they are not contiguous with
eachother. First we have the malloc mapping area, then another
unrelated region, then the vmalloc region.
This "another unrelated region" is where the firmware is mapped.
If the lazy TLB flushing logic in the vmalloc code triggers after
we've had both a module unload and a vfree or similar, it will pass an
address range that goes from somewhere inside the malloc region to
somewhere inside the vmalloc region, and thus covering the
openfirmware area entirely.
The sparc64 kernel learns about openfirmware's dynamic mappings in
this region early in the boot, and then services TLB misses in this
area. But openfirmware has some locked TLB entries which are not
mentioned in those dynamic mappings and we should thus not disturb
them.
These huge lazy TLB flush ranges causes those openfirmware locked TLB
entries to be removed, resulting in all kinds of problems including
hard hangs and crashes during reboot/reset.
Besides causing problems like this, such huge TLB flush ranges are
also incredibly inefficient. A plea has been made with the author of
the VMAP lazy TLB flushing code, but for now we'll put a safety guard
into our flush_tlb_kernel_range() implementation.
Since the implementation has become non-trivial, stop defining it as a
macro and instead make it a function in a C source file.
Signed-off-by: David S. Miller <davem@davemloft.net>
The assumption was that update_mmu_cache() (and the equivalent for PMDs) would
only be called when the PTE being installed will be accessible by the user.
This is not true for code paths originating from remove_migration_pte().
There are dire consequences for placing a non-valid PTE into the TSB. The TLB
miss frramework assumes thatwhen a TSB entry matches we can just load it into
the TLB and return from the TLB miss trap.
So if a non-valid PTE is in there, we will deadlock taking the TLB miss over
and over, never satisfying the miss.
Just exit early from update_mmu_cache() and friends in this situation.
Based upon a report and patch from Christopher Alexander Tobias Schulze.
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds sparc RAM to /proc/iomem. It also identifies the
code, data and bss regions of the kernel.
Signed-off-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix following warnings:
init_64.c:798:5: warning: symbol 'numa_cpu_lookup_table' was not declared. Should it be static?
init_64.c:799:11: warning: symbol 'numa_cpumask_lookup_table' was not declared. Should it be static?
The warnings were present with an allnoconfig
Fix so the variables are only declared if CONFIG_NEED_MULTIPLE_NODES is defined.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix following warnings:
init_64.c:191:10: warning: symbol 'dcpage_flushes' was not declared. Should it be static?
init_64.c:193:10: warning: symbol 'dcpage_flushes_xcall' was not declared. Should it be static?
Add extern declaration to asm/setup.h and drop local declaration in smp_64.h
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
After merging the final tree, today's linux-next build (sparc64 defconfig)
failed like this:
arch/sparc/mm/init_64.c: In function 'pte_alloc_one':
arch/sparc/mm/init_64.c:2568:9: error: unused variable 'pte' [-Werror=unused-variable]
Caused by the merge between commit 37b3a8ff3e ("sparc64: Move from 4MB
to 8MB huge pages") and commit 1ae9ae5f7d ("sparc: handle
pgtable_page_ctor() fail") (I had the following merge fix in linux-next,
but it didn't seem to propagate upstream - may have forgotten to point it
out :-().
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull sparc update from David Miller:
1) Implement support for up to 47-bit physical addresses on sparc64.
2) Support HAVE_CONTEXT_TRACKING on sparc64, from Kirill Tkhai.
3) Fix Simba bridge window calculations, from Kjetil Oftedal.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-next:
sparc64: Implement HAVE_CONTEXT_TRACKING
sparc64: Add self-IPI support for smp_send_reschedule()
sparc: PCI: Fix incorrect address calculation of PCI Bridge windows on Simba-bridges
sparc64: Encode huge PMDs using PTE encoding.
sparc64: Move to 64-bit PGDs and PMDs.
sparc64: Move from 4MB to 8MB huge pages.
sparc64: Make PAGE_OFFSET variable.
sparc64: Fix inconsistent max-physical-address defines.
sparc64: Document the shift counts used to validate linear kernel addresses.
sparc64: Define PAGE_OFFSET in terms of physical address bits.
sparc64: Use PAGE_OFFSET instead of a magic constant.
sparc64: Clean up 64-bit mmap exclusion defines.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now that we have 64-bits for PMDs we can stop using special encodings
for the huge PMD values, and just put real PTEs in there.
We allocate a _PAGE_PMD_HUGE bit to distinguish between plain PMDs and
huge ones. It is the same for both 4U and 4V PTE layouts.
We also use _PAGE_SPECIAL to indicate the splitting state, since a
huge PMD cannot also be special.
All of the PMD --> PTE translation code disappears, and most of the
huge PMD bit modifications and tests just degenerate into the PTE
operations. In particular USER_PGTABLE_CHECK_PMD_HUGE becomes
trivial.
As a side effect, normal PMDs don't shift the physical address around.
This also speeds up the page table walks in the TLB miss paths since
they don't have to do the shifts any more.
Another non-trivial aspect is that pte_modify() has to be changed
to preserve the _PAGE_PMD_HUGE bits as well as the page size field
of the pte.
Signed-off-by: David S. Miller <davem@davemloft.net>
To make the page tables compact, we were using 32-bit PGDs and PMDs.
We only had to support <= 43 bits of physical addresses so this was
quite feasible.
In order to support larger physical addresses we have to move to
64-bit PGDs and PMDs.
Most of the changes are straight-forward:
1) {pgd,pmd}_t --> unsigned long
2) Anything that tries to use plain "unsigned int" types with pgd/pmd
values needs to be adjusted. In particular things like "0U" become
"0UL".
3) {PGDIR,PMD}_BITS decrease by one.
4) In the assembler page table walkers, use "ldxa" instead of "lduwa"
and adjust the low bit masks to clear out the low 3 bits instead of
just the low 2 bits during pgd/pmd address formation.
Also, use PTRS_PER_PGD and PTRS_PER_PMD in the sizing of the
swapper_{pg_dir,low_pmd_dir} arrays.
This patch does not try to take advantage of having 64-bits in the
PMDs to simplify the hugepage code, that will come in a subsequent
change.
Signed-off-by: David S. Miller <davem@davemloft.net>
The impetus for this is that we would like to move to 64-bit PMDs and
PGDs, but that would result in only supporting a 42-bit address space
with the current page table layout. It'd be nice to support at least
43-bits.
The reason we'd end up with only 42-bits after making PMDs and PGDs
64-bit is that we only use half-page sized PTE tables in order to make
PMDs line up to 4MB, the hardware huge page size we use.
So what we do here is we make huge pages 8MB, and fabricate them using
4MB hw TLB entries.
Facilitate this by providing a "REAL_HPAGE_SHIFT" which is used in
places that really need to operate on hardware 4MB pages.
Use full pages (512 entries) for PTE tables, and adjust PMD_SHIFT,
PGD_SHIFT, and the build time CPP test as needed. Use a CPP test to
make sure REAL_HPAGE_SHIFT and the _PAGE_SZHUGE_* we use match up.
This makes the pgtable cache completely unused, so remove the code
managing it and the state used in mm_context_t. Now we have less
spinlocks taken in the page table allocation path.
The technique we use to fabricate the 8MB pages is to transfer bit 22
from the missing virtual address into the PTEs physical address field.
That takes care of the transparent huge pages case.
For hugetlb, we fill things in at the PTE level and that code already
puts the sub huge page physical bits into the PTEs, based upon the
offset, so there is nothing special we need to do. It all just works
out.
So, a small amount of complexity in the THP case, but this code is
about to get much simpler when we move the 64-bit PMDs as we can move
away from the fancy 32-bit huge PMD encoding and just put a real PTE
value in there.
With bug fixes and help from Bob Picco.
Signed-off-by: David S. Miller <davem@davemloft.net>
Choose PAGE_OFFSET dynamically based upon cpu type.
Original UltraSPARC-I (spitfire) chips only supported a 44-bit
virtual address space.
Newer chips (T4 and later) support 52-bit virtual addresses
and up to 47-bits of physical memory space.
Therefore we have to adjust PAGE_SIZE dynamically based upon
the capabilities of the chip.
Note that this change alone does not allow us to support > 43-bit
physical memory, to do that we need to re-arrange our page table
support. The current encodings of the pmd_t and pgd_t pointers
restricts us to "32 + 11" == 43 bits.
This change can waste quite a bit of memory for the various tables.
In particular, a future change should work to size and allocate
kern_linear_bitmap[] and sparc64_valid_addr_bitmap[] dynamically.
This isn't easy as we really cannot take a TLB miss when accessing
kern_linear_bitmap[]. We'd have to lock it into the TLB or similar.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Bob Picco <bob.picco@oracle.com>
Nitin Gupta