When we map memory at boot we print out the ranges of real addresses
that we mapped and the page size that was used.
Currently it's a bit ugly:
Mapped range 0x0 - 0x2000000000 with 0x40000000
Mapped range 0x200000000000 - 0x202000000000 with 0x40000000
Pad the addresses so they line up, and print the page size using
actual units, eg:
Mapped 0x0000000000000000-0x0000000001200000 with 64.0 KiB pages
Mapped 0x0000000001200000-0x0000000040000000 with 2.00 MiB pages
Mapped 0x0000000040000000-0x0000000100000000 with 1.00 GiB pages
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
It's too big to be inline, there is no reason to keep it
that way.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[mpe: Rework to incorporate the comment changes via fixes branch]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Instead of comparing the whole CPU mask every time, let's
keep a counter of how many bits are set in the mask. Thus
testing for a local mm only requires testing if that counter
is 1 and the current CPU bit is set in the mask.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We open-code testing for the mm being local to the current CPU
in a few places. Use our existing helper instead.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There's a non-trivial dependency between some commits we want to put in
next and the KVM prefetch work around that went into fixes. So merge
fixes into next.
We use mm cpumask for serializing against lockless page table walk.
Anybody who is doing a lockless page table walk is expected to disable
irq and only cpus in mm cpumask is expected do the lockless walk. This
ensure that a THP split can send IPI to only cpus in the mm cpumask,
to make sure there are no parallel lockless page table walk.
Add the CAPI fault handling cpu to the mm cpumask so that we can do
the lockless page table walk while inserting hash page table entries.
Reviewed-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Now that we made sure that lockless walk of linux page table is mostly
limitted to current task(current->mm->pgdir) we can update the THP
update sequence to only send IPI to CPUs on which this task has run.
This helps in reducing the IPI overload on systems with large number
of CPUs.
WRT kvm even though kvm is walking page table with vpc->arch.pgdir,
it is done only on secondary CPUs and in that case we have primary CPU
added to task's mm cpumask. Sending an IPI to primary will force the
secondary to do a vm exit and hence this mm cpumask usage is safe
here.
WRT CAPI, we still end up walking linux page table with capi context
MM. For now the pte lookup serialization sends an IPI to all CPUs in
CPI is in use. We can further improve this by adding the CAPI
interrupt handling CPU to task mm cpumask. That will be done in a
later patch.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Bring in the commit to rename find_linux_pte_or_hugepte() which touches
arch and KVM code, and might need to be merged with the kvmppc tree to
avoid conflicts.
Add newer helpers to make the function usage simpler. It is always
recommended to use find_current_mm_pte() for walking the page table.
If we cannot use find_current_mm_pte(), it should be documented why
the said usage of __find_linux_pte() is safe against a parallel THP
split.
For now we have KVM code using __find_linux_pte(). This is because kvm
code ends up calling __find_linux_pte() in real mode with MSR_EE=0 but
with PACA soft_enabled = 1. We may want to fix that later and make
sure we keep the MSR_EE and PACA soft_enabled in sync. When we do that
we can switch kvm to use find_linux_pte().
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
With commit aa888a7497 ("hugetlb: support larger than MAX_ORDER") we added
support for allocating gigantic hugepages via kernel command line. Switch
ppc64 arch specific code to use that.
W.r.t FSL support, we now limit our allocation range using BOOTMEM_ALLOC_ACCESSIBLE.
We use the kernel command line to do reservation of hugetlb pages on powernv
platforms. On pseries hash mmu mode the supported gigantic huge page size is
16GB and that can only be allocated with hypervisor assist. For pseries the
command line option doesn't do the allocation. Instead pseries does gigantic
hugepage allocation based on hypervisor hint that is specified via
"ibm,expected#pages" property of the memory node.
Cc: Scott Wood <oss@buserror.net>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
gup_hugepte() checks if pages are present and readable, and
when 'write' is set, also checks if the pages are writable.
Initially this was done by checking if _PAGE_PRESENT and
_PAGE_READ were set. In addition, _PAGE_WRITE was verified for write
accesses.
The problem is that we have to handle the three following cases:
1/ The target defines __PAGE_READ and __PAGE_WRITE
2/ The target defines __PAGE_RW
3/ The target defines __PAGE_RO
In case 1/, this is obvious
In case 2/, __PAGE_READ is defined as 0 and __PAGE_WRITE as __PAGE_RW
so it works as well.
But in case 3, __PAGE_RW is defined as 0, which means __PAGE_WRITE is 0
and then the test returns true (page writable) in all cases.
A first correction was attempted in commit 6b8cb66a6a ("powerpc: Fix
usage of _PAGE_RO in hugepage"), but that fix is wrong:
instead of checking that the page is writable when write is requested,
it checks that the page is NOT writable when write is NOT requested.
This patch adds a new pte_read() helper to check whether a page is
readable or not. This avoids handling all possible cases in
gup_hugepte().
Then gup_hugepte() is modified to use pte_present(), pte_read()
and pte_write() instead of the raw flags.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
__set_fixmap() uses __fix_to_virt() then does the boundary checks
by it self. Instead, we can use fix_to_virt() which does the
verification at build time. For this, we need to use it inline
so that GCC can see the real value of idx at buildtime.
In the meantime, we remove the 'fixmaps' variable.
This variable is set but has never been used from the beginning
(commit 2c419bdeca ("[POWERPC] Port fixmap from x86 and use
for kmap_atomic"))
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
get_pteptr() and __mapin_ram_chunk() are only used locally,
so define them static
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This patch implements STRICT_KERNEL_RWX on PPC32.
As for CONFIG_DEBUG_PAGEALLOC, it deactivates BAT and LTLB mappings
in order to allow page protection setup at the level of each page.
As BAT/LTLB mappings are deactivated, there might be a performance
impact.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
As seen below, allthough the init sections have been freed, the
associated memory area is still marked as executable in the
page tables.
~ dmesg
[ 5.860093] Freeing unused kernel memory: 592K (c0570000 - c0604000)
~ cat /sys/kernel/debug/kernel_page_tables
---[ Start of kernel VM ]---
0xc0000000-0xc0497fff 4704K rw X present dirty accessed shared
0xc0498000-0xc056ffff 864K rw present dirty accessed shared
0xc0570000-0xc059ffff 192K rw X present dirty accessed shared
0xc05a0000-0xc7ffffff 125312K rw present dirty accessed shared
---[ vmalloc() Area ]---
This patch fixes that.
The implementation is done by reusing the change_page_attr()
function implemented for CONFIG_DEBUG_PAGEALLOC
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
__change_page_attr() uses flush_tlb_page().
flush_tlb_page() uses tlbie instruction, which also invalidates
pinned TLBs, which is not what we expect.
This patch modifies the implementation to use flush_tlb_kernel_range()
instead. This will make use of tlbia which will preserve pinned TLBs.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
setup_initial_memory_limit() is only called during init.
mmu_patch_cmp_limit() is only called from 8xx_mmu.c
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
As stated in a comment in head_8xx.S, today we "Always pin the first
8 MB ITLB to prevent ITLB misses while mucking around with SRR0/SRR1
in asm".
This issue has just been cleared by the preceding patch, therefore
we can make this pinning optional (on by default) and independent
of DATA pinning.
This patch also makes pinning of IMMR independent of pinning of DATA.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
On the 8xx, the RAM mapped with LTLBs must be seen as block mapped,
just like areas mapped with BATs on standard PPC32.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Use symbolic names for DSISR bits in DSI
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Two config options exist to define powerpc MPC8xx:
* CONFIG_PPC_8xx
* CONFIG_8xx
arch/powerpc/platforms/Kconfig.cputype has contained the following
comment about CONFIG_8xx item for some years:
"# this is temp to handle compat with arch=ppc"
arch/powerpc is now the only place with remaining use of
CONFIG_8xx: get rid of them.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The host process table base is stored in the partition table by calling
the function native_register_process_table(). Currently this just sets
the entry in memory and is missing a subsequent cache invalidation
instruction. Any update to the partition table should be followed by a
cache invalidation instruction specifying invalidation of the caching of
any partition table entries (RIC = 2, PRS = 0).
We already have a function to update the partition table with the
required cache invalidation instructions - mmu_partition_table_set_entry().
Update the native_register_process_table() function to call
mmu_partition_table_set_entry(), this ensures all appropriate
invalidation will be performed.
Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
[mpe: Use a local for patb0 to clean it up slightly]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
On 64-bit book3s, with the hash MMU, we currently define the kernel
virtual space (vmalloc, ioremap etc.), to be 16T in size. This is a
leftover from pre v3.7 when our user VM was also 16T.
Of that 16T we split it 50/50, with half used for PCI IO and ioremap
and the other 8T for vmalloc.
We never bothered to make it any bigger because 8T of vmalloc ought to
be enough for anybody. But it turns out that's not true, the per cpu
allocator wants large amounts of vmalloc space, not to make large
allocations, but to allow a large stride between allocations, because
we use pcpu_embed_first_chunk().
With a bit of juggling we can increase the entire kernel virtual space
to 64T. The only real complication is the check of the address in the
SLB miss handler, see the comment in the code.
Although we could continue to split virtual space 50/50 as we do now,
no one seems to be running out of PCI IO or ioremap space. So instead
keep that as 8T, and use the remaining 56T for vmalloc.
In future we should be able to increase the kernel virtual space to
512T, the code already supports that, it just needs testing on older
hardware.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
There is a comment in slb_allocate() referring to the load of
paca->vmalloc_sllp, but it's several lines prior in the assembly.
We're about to change this code, and we want to add another comment,
so move the comment immediately prior to the instruction it's talking
about.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Currently KERN_IO_START is defined as:
#define KERN_IO_START (KERN_VIRT_START + (KERN_VIRT_SIZE >> 1))
Although it looks like a constant, both the components are actually
variables, to allow us to have a different value between Radix and
Hash with a single kernel.
However that still requires both Radix and Hash to place the kernel IO
region at the same location relative to the start and end of the
kernel virtual region (namely 1/2 way through it), and we'd like to
change that.
So split KERN_IO_START out into its own variable, and initialise it
for Radix and Hash. In the medium term we should be able to
reconsolidate this, by doing a more involved rearrangement of the
location of the regions.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We have a whole pile of unused code to maintain the ACOP register,
allocate coprocessor PIDs and handle ACOP faults. This mechanism
was used for the HFI adapter on POWER7 which is dead and gone and
whose driver never went upstream. It was used on some A2 core based
stuff that also never saw the light of day.
Take out all that code.
There is still some POWER8 coprocessor code that uses icswx but it's
kernel only and thus doesn't use any of that infrastructure.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When hitting below a VM_GROWSDOWN vma (typically growing the stack),
we check whether it's a valid stack-growing instruction and we
check the distance to GPR1. This is largely open coded with lots
of comments, so move it out to a helper.
While at it, make store_update_sp a boolean.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
If the first iteration returns VM_FAULT_MAJOR but the second
one doesn't, we fail to account the fault as a major fault.
This fixes it and brings the code in line with x86.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Move out the code that sets FAULT_FLAG_WRITE so the block that check
access permissions can be extracted. While at it also set
FAULT_FLAG_INSTRUCTION which will be used for protection keys.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Do the check before we re-enable interrupts and clean the code
up a bit.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This has a page of comment explaining what's going on right in
the middle of do_page_fault() which makes things a bit hard to
follow. Move it to a helper instead. Also do the test earlier
as there's no point waiting until after we found the VMA.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
No need to break those lines, they aren't that long
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
It makes do_page_fault() more readable. No functional change.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
First, handle the normal retry failure in do_page_fault itself,
since it's a simple return statement. That allows us to remove
the "continue" special return code from mm_fault_error().
Once that's done, we can have an implementation much closer to
x86 where we only call mm_fault_error() if VM_FAULT_ERROR is set
and directly return.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Instead of goto labels, instead call those functions and return.
This gets us closer to x86 and allows us to shring do_page_fault()
even more.
The main difference with x86 is that those function return a value
which we then return from do_page_fault(). That value is our
return value from do_page_fault() which we use to generate
kernel faults.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We currently test for is_exec and DSISR_PROTFAULT but that doesn't
make sense as this is the wrong error bit to test for an execute
permission failure.
In fact, we had code that would return early if we had an exec
fault in kernel mode so I think that was just dead code anyway.
Finally the location of that test is awkward and prevents further
simplifications.
So instead move that test into a helper along with the existing
early test for kernel exec faults and out of range accesses,
and put it all in a "bad_kernel_fault()" helper. While at it
test the correct error bits.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Now that we moved the exception state handling to a wrapper, we can
just directly return rather than "goto bail"
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
A bad page fault is when the HW signals an error such as a bad
copy/paste, an AMO error, or some other type of error that will
not be fixed by updating the PTE.
Use a helper page_fault_is_bad() to check for bad page faults thus
removing the per-processor family open-coding in __do_page_fault()
and trigger a SIGBUS rather than a SIGSEGV which is more appropriate.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There's no point looking for the VMA etc.. when we already know
we are going to fail.
This adds some code to set "code" for the si_code but that will
be gone in subsequent patches.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Define a common page_fault_is_write() helper and use it
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
On legacy 6xx 32-bit procesors, we checked for the DABR match bit
in DSISR from do_page_fault(), in the middle of a pile of ifdef's
because all other CPU types do it in assembly prior to calling
do_page_fault. Fix that.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[mpe: Add #ifdef CONFIG_6xx]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
By filtering the relevant SRR1 bits in the assembly rather than
in do_page_fault() itself, we avoid a conditional branch (since we
already come from different path for data and instruction faults).
This will allow more simplifications later
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This will allow simplifying the returns from do_page_fault
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We do that because it's used by THP pmd collapsing, so use
instead a dedicated flush function.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
At the moment we have to rather sub-optimal flushing behaviours:
- flush_tlb_mm() will flush the PWC which is unnecessary (for example
when doing a fork)
- A large unmap will call flush_tlb_pwc() multiple times causing us
to perform that fairly expensive operation repeatedly. This happens
often in batches of 3 on every new process.
So we change flush_tlb_mm() to only flush the TLB, and we use the
existing "need_flush_all" flag in struct mmu_gather to indicate
that the PWC needs flushing.
Unfortunately, flush_tlb_range() still needs to do a full flush
for now as it's used by the THP collapsing. We will fix that later.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Michael Ellerman