Add logic to the power8 PMU code to support EBB. Future processors would
also be expected to implement similar constraints. At that time we could
possibly factor these out into common code.
Finally mark the power8 PMU as supporting EBB, which is the actual
enable switch which allows EBBs to be configured.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Add support for EBB (Event Based Branches) on 64-bit book3s. See the
included documentation for more details.
EBBs are a feature which allows the hardware to branch directly to a
specified user space address when a PMU event overflows. This can be
used by programs for self-monitoring with no kernel involvement in the
inner loop.
Most of the logic is in the generic book3s code, primarily to avoid a
proliferation of PMU callbacks.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In commit 59affcd "Context switch more PMU related SPRs" I added more
PMU SPRs to thread_struct, later modified in commit b11ae95. To add
insult to injury it turns out we don't need to switch MMCRA as it's
only user readable, and the value is recomputed by the PMU code.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In power_pmu_enable() we still enable the PMU even if we have zero
events. This should have no effect but doesn't make much sense. Instead
just return after telling the hypervisor that we are not using the PMCs.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In power_pmu_enable() we can use the existing out label to reduce the
number of return paths.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On Power8 we can freeze PMC5 and 6 if we're not using them. Normally they
run all the time.
As noticed by Anshuman, we should unfreeze them when we disable the PMU
as there are legacy tools which expect them to run all the time.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In pmu_disable() we disable the PMU by setting the FC (Freeze Counters)
bit in MMCR0. In order to do this we have to read/modify/write MMCR0.
It's possible that we read a value from MMCR0 which has PMAO (PMU Alert
Occurred) set. When we write that value back it will cause an interrupt
to occur. We will then end up in the PMU interrupt handler even though
we are supposed to have just disabled the PMU.
We can avoid this by making sure we never write PMAO back. We should not
lose interrupts because when the PMU is re-enabled the overflowed values
will cause another interrupt.
We also reorder the clearing of SAMPLE_ENABLE so that is done after the
PMU is frozen. Otherwise there is a small window between the clearing of
SAMPLE_ENABLE and the setting of FC where we could take an interrupt and
incorrectly see SAMPLE_ENABLE not set. This would for example change the
logic in perf_read_regs().
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
A mistake we have made in the past is that we pull out the fields we
need from the event code, but don't check that there are no unknown bits
set. This means that we can't ever assign meaning to those unknown bits
in future.
Although we have once again failed to do this at release, it is still
early days for Power8 so I think we can still slip this in and get away
with it.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Similar to the facility unavailble exception, except the facilities are
controlled by HFSCR.
Adapt the facility_unavailable_exception() so it can be called for
either the regular or Hypervisor facility unavailable exceptions.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The exception at 0xf60 is not the TM (Transactional Memory) unavailable
exception, it is the "Facility Unavailable Exception", rename it as
such.
Flesh out the handler to acknowledge the fact that it can be called for
many reasons, one of which is TM being unavailable.
Use STD_EXCEPTION_COMMON() for the exception body, for some reason we
had it open-coded, I've checked the generated code is identical.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.10]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
KVMTEST is a macro which checks whether we are taking an exception from
guest context, if so we branch out of line and eventually call into the
KVM code to handle the switch.
When running real guests on bare metal (HV KVM) the hardware ensures
that we never take a relocation on exception when transitioning from
guest to host. For PR KVM we disable relocation on exceptions ourself in
kvmppc_core_init_vm(), as of commit a413f47 "Disable relocation on
exceptions whenever PR KVM is active".
So convert all the RELON macros to use NOTEST, and drop the remaining
KVM_HANDLER() definitions we have for 0xe40 and 0xe80.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.9+]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We have relocation on exception handlers defined for h_data_storage and
h_instr_storage. However we will never take relocation on exceptions for
these because they can only come from a guest, and we never take
relocation on exceptions when we transition from guest to host.
We also have a handler for hmi_exception (Hypervisor Maintenance) which
is defined in the architecture to never be delivered with relocation on,
see see v2.07 Book III-S section 6.5.
So remove the handlers, leaving a branch to self just to be double extra
paranoid.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
CC: <stable@vger.kernel.org> [v3.9+]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The topology update code that updates the cpu node registration in sysfs
should not be called while in stop_machine(). The register/unregister
calls take a lock and may sleep.
This patch moves these calls outside of the call to stop_machine().
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
the smp_release_cpus is a normal funciton and called in normal environments,
but it calls the __initdata spinning_secondaries.
need modify spinning_secondaries to match smp_release_cpus.
the related warning:
(the linker report boot_paca.33377, but it should be spinning_secondaries)
-----------------------------------------------------------------------------
WARNING: arch/powerpc/kernel/built-in.o(.text+0x23176): Section mismatch in reference from the function .smp_release_cpus() to the variable .init.data:boot_paca.33377
The function .smp_release_cpus() references
the variable __initdata boot_paca.33377.
This is often because .smp_release_cpus lacks a __initdata
annotation or the annotation of boot_paca.33377 is wrong.
WARNING: arch/powerpc/kernel/built-in.o(.text+0x231fe): Section mismatch in reference from the function .smp_release_cpus() to the variable .init.data:boot_paca.33377
The function .smp_release_cpus() references
the variable __initdata boot_paca.33377.
This is often because .smp_release_cpus lacks a __initdata
annotation or the annotation of boot_paca.33377 is wrong.
-----------------------------------------------------------------------------
Signed-off-by: Chen Gang <gang.chen@asianux.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
A few new i2c-drivers came into the kernel which clear the clientdata-pointer
on exit or error. This is obsolete meanwhile, the core will do it.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
When error occurs, need return the related error code to let upper
caller know about it.
ppc_md.nvram_size() can return the error code (e.g. core99_nvram_size()
in 'arch/powerpc/platforms/powermac/nvram.c').
Also set ret value when only need it, so can save structions for normal
cases.
Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
cuda_init_via() is called from find_via_cuda() only, which is __init.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The __cpuinit type of throwaway sections might have made sense
some time ago when RAM was more constrained, but now the savings
do not offset the cost and complications. For example, the fix in
commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time")
is a good example of the nasty type of bugs that can be created
with improper use of the various __init prefixes.
After a discussion on LKML[1] it was decided that cpuinit should go
the way of devinit and be phased out. Once all the users are gone,
we can then finally remove the macros themselves from linux/init.h.
This removes all the powerpc uses of the __cpuinit macros. There
are no __CPUINIT users in assembly files in powerpc.
[1] https://lkml.org/lkml/2013/5/20/589
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Josh Boyer <jwboyer@gmail.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This typedef is unnecessary and should just be removed.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This typedef is unnecessary and should just be removed.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
pci_iommu_init() and pci_direct_iommu_init() are not referenced anywhere,
so remove them.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
For an unknown relocation type since the value of r4 is just the 8bit
relocation type, the sum of r4 and r7 may yield an invalid memory
address. For example:
In normal case:
r4 = c00xxxxx
r7 = 40000000
r4 + r7 = 000xxxxx
For an unknown relocation type:
r4 = 000000xx
r7 = 40000000
r4 + r7 = 400000xx
400000xx is an invalid memory address for a board which has just
512M memory.
And for operations such as dcbst or icbi may cause bus error for an
invalid memory address on some platforms and then cause the board
reset. So we should skip the flush/invalidate the d/icache for
an unknown relocation type.
Signed-off-by: Kevin Hao <haokexin@gmail.com>
Acked-by: Suzuki K. Poulose <suzuki@in.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
With pm81/pm91/pm121, when the overtemperature state is entered, and
when it remains on after skipped ticks, the driver will try to leave
it too soon (immediately on the next tick). This is because the active
FAILURE_OVERTEMP state is not visible in "new_failure" variable of the
current tick. Furthermore, the driver will keep trying to clear condition
in subsequent ticks as FAILURE_OVERTEMP remains set in the "last_failure"
variable. These will start to trigger WARNINGS from windfarm core:
[ 100.082735] windfarm: Clamping CPU frequency to minimum !
[ 100.108132] windfarm: Overtemp condition detected !
[ 101.952908] windfarm: Overtemp condition cleared !
[...]
[ 102.980388] WARNING: at drivers/macintosh/windfarm_core.c:463
[...]
[ 103.982227] WARNING: at drivers/macintosh/windfarm_core.c:463
[...]
[ 105.030494] WARNING: at drivers/macintosh/windfarm_core.c:463
[...]
[ 105.973666] WARNING: at drivers/macintosh/windfarm_core.c:463
[...]
[ 106.977913] WARNING: at drivers/macintosh/windfarm_core.c:463
Fix by adding a helper global variable. We leave the overtemp state only
after all failure bits have been cleared.
I saw this error on iMac G5 iSight (pm121). Also pm81/pm91 are fixed
based on the observation that these are almost identical/copy-pasted code.
Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently, we're using the combo (PCI bus + devfn) in the PCI
config accessors and PCI config accessors in EEH depends on them.
However, it's not safe to refer the PCI bus which might have been
removed during hotplug. So we're using device node in the PCI
config accessors and the corresponding backends just reuse them.
The patch also fix one potential risk: We possiblly have frozen
PE during the early PCI probe time, but we haven't setup the PE
mapping yet. So the errors should be counted to PE#0.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The patch is for avoiding following build warnings:
The function .pnv_pci_ioda_fixup() references
the function __init .eeh_init().
This is often because .pnv_pci_ioda_fixup lacks a __init
The function .pnv_pci_ioda_fixup() references
the function __init .eeh_addr_cache_build().
This is often because .pnv_pci_ioda_fixup lacks a __init
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We needn't the the whole backtrace other than one-line message in
the error reporting interrupt handler. For errors triggered by
access PCI config space or MMIO, we replace "WARN(1, ...)" with
pr_err() and dump_stack(). The patch also adds more output messages
to indicate what EEH core is doing. Besides, some printk() are
replaced with pr_warning().
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On the PowerNV platform, the EEH address cache isn't built correctly
because we skipped the EEH devices without binding PE. The patch
fixes that.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We have 2 fields in "struct pnv_phb" to trace the states. The patch
replace the fields with one and introduces flags for that. The patch
doesn't impact the logic.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
After reset (e.g. complete reset) in order to bring the fenced PHB
back, the PCIe link might not be ready yet. The patch intends to
make sure the PCIe link is ready before accessing its subordinate
PCI devices. The patch also fixes that wrong values restored to
PCI_COMMAND register for PCI bridges.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
When the PHB is fenced or dead, it's pointless to collect the data
from PCI config space of subordinate PCI devices since it should
return 0xFF's. The patch also fixes overwritten buffer while getting
PCI config data.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
When we treclaim and trecheckpoint there's an unavoidable period when r1
will not be a valid kernel stack pointer.
This patch clears the MSR recoverable interrupt (RI) bit over these
regions to indicate we have an invalid kernel stack pointer.
For treclaim, the region over which we clear MSR RI is larger than
required to avoid the need for an extra costly mtmsrd.
Thanks to Paulus for suggesting this change.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
String instruction emulation would erroneously result in a segfault if
the upper bits of the EA are set and is so high that it fails access
check. Truncate the EA to 32 bits if the process is 32-bit.
Signed-off-by: James Yang <James.Yang@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently, HPAGE_PMD_* constans rely on PMD_SHIFT regardless of
CONFIG_TRANSPARENT_HUGEPAGE. PMD_SHIFT is not defined everywhere (e.g.
arm nommu case).
It means we can't use anything like this in generic code:
if (PageTransHuge(page))
zero_huge_user(page, 0, HPAGE_PMD_SIZE);
else
clear_highpage(page);
For !THP case, PageTransHuge() is 0 and compiler can eliminate
zero_huge_user() call. But it still need to be valid C expression, means
HPAGE_PMD_SIZE has to expand to something compiler can understand.
Previously, HPAGE_PMD_* were defined to BUILD_BUG() for !THP. Let's come
back to it.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
To replace down() with down_interrutible() to avoid following
warning:
[c00000007ba7b710] [c000000000014410] .__switch_to+0x1b0/0x380
[c00000007ba7b7c0] [c0000000007b408c] .__schedule+0x3ec/0x970
[c00000007ba7ba50] [c0000000007b1f24] .schedule_timeout+0x1a4/0x2b0
[c00000007ba7bb30] [c0000000007b34a4] .__down+0xa4/0x104
[c00000007ba7bbf0] [c0000000000b9230] .down+0x60/0x70
[c00000007ba7bc80] [c0000000000336d0] .eeh_event_handler+0x70/0x190
[c00000007ba7bd30] [c0000000000b1a58] .kthread+0xe8/0xf0
[c00000007ba7be30] [c00000000000a05c] .ret_from_kernel_thread+0x5c/0x8
This also avoids keeping the load average up while doing nothing.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Originally, eeh_mutex was introduced to protect the PE hierarchy
tree and the attached EEH devices because EEH core was possiblly
running with multiple threads to access the PE hierarchy tree.
However, we now have only one kthread in EEH core. So we needn't
the eeh_mutex and just remove it.
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Building with CONFIG_TRANSPARENT_HUGEPAGE disabled causes the following
build wearnings;
powerpc/arch/powerpc/include/asm/mmu-hash64.h: In function ‘__hash_page_thp’:
powerpc/arch/powerpc/include/asm/mmu-hash64.h:354: warning: no return statement in function returning non-void
This patch adds a return -1 to the static inline for __hash_page_thp()
to correct the warnings.
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Since now we have pstore support for nvram in pseries, enable it
in the default config. With this config option enabled, pstore
infra-structure will be used to read/write the messages from/to nvram.
Signed-off-by: Aruna Balakrishnaiah <aruna@linux.vnet.ibm.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In 9422de3 "powerpc: Hardware breakpoints rewrite to handle non DABR breakpoint
registers" we changed the way we mark extraneous irqs with this:
- info->extraneous_interrupt = !((bp->attr.bp_addr <= dar) &&
- (dar - bp->attr.bp_addr < bp->attr.bp_len));
+ if (!((bp->attr.bp_addr <= dar) &&
+ (dar - bp->attr.bp_addr < bp->attr.bp_len)))
+ info->type |= HW_BRK_TYPE_EXTRANEOUS_IRQ;
Unfortunately this is bogus as it never clears extraneous IRQ if it's already
set.
This correctly clears extraneous IRQ before possibly setting it.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Reported-by: Edjunior Barbosa Machado <emachado@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The smallest match region for both the DABR and DAWR is 8 bytes, so the
kernel needs to filter matches when users want to look at regions smaller than
this.
Currently we set the length of PPC_BREAKPOINT_MODE_EXACT breakpoints to 8.
This is wrong as in exact mode we should only match on 1 address, hence the
length should be 1.
This ensures that the kernel will filter out any exact mode hardware breakpoint
matches on any addresses other than the requested one.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Reported-by: Edjunior Barbosa Machado <emachado@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Hugepage invalidate involves invalidating multiple hpte entries.
Optimize the operation using H_BULK_REMOVE on lpar platforms.
On native, reduce the number of tlb flush.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We enable only if the we support 16MB page size.
Reviewed-by: David Gibson <dwg@au1.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We find all the overlapping vma and mark them such that we don't allocate
hugepage in that range. Also we split existing huge page so that the
normal page hash can be invalidated and new page faulted in with new
protection bits.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
With THP we set pmd to none, before we do pte_clear. Hence we can't
walk page table to get the pte lock ptr and verify whether it is locked.
THP do take pte lock before calling pte_clear. So we don't change the locking
rules here. It is that we can't use page table walking to check whether
pte locks are held with THP.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
GCC is very likely to read the pagetables just once and cache them in
the local stack or in a register, but it is can also decide to re-read
the pagetables. The problem is that the pagetable in those places can
change from under gcc.
With THP/hugetlbfs the pmd (and pgd for hugetlbfs giga pages) can
change under gup_fast. The pages won't be freed untill we finish
gup fast because we have irq disabled and we free these pages via
rcu callback.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We need to have irqs disabled to handle all the possible parallel update for
linux page table without holding locks.
Events that we are intersted in while walking page tables are
1) Page fault
2) umap
3) THP split
4) THP collapse
A) local_irq_disabled:
------------------------
1) page fault:
A none to valid transition via page fault is not an issue because we
would either see a none or valid. If it is none, we would error out
the page table walk. We may need to use on stack values when checking for
type of page table elements, because if we do
if (!is_hugepd()) {
if (!pmd_none() {
if (pmd_bad() {
We could take that bad condition because the pmd got converted to a hugepd
after the !is_hugepd check via a hugetlb fault.
The right way would be to check for pmd_none higher up or use on stack value.
2) A valid to none conversion via unmap:
We can safely walk the upper level table, because we don't remove the the
page table entries until rcu grace period. So even if we followed a
wrong pointer we still have the pointer valid till the grace period.
A PTE pointer returned need to be atomically checked for _PAGE_PRESENT and
_PAGE_BUSY. A valid pointer returned could becoming none later. To prevent
pte_clear we take _PAGE_BUSY.
3) THP split:
A valid transparent hugepage is converted to nomal page. Before we split we
do pmd_splitting_flush, which sets the hugepage PTE to _PAGE_SPLITTING
So when walking page table we need to check for pmd_trans_splitting and
handle that. The pte returned should also need to be checked for
_PAGE_SPLITTING before setting _PAGE_BUSY similar to _PAGE_PRESENT. We save
the value of PTE on stack and check for the flag in the local pte value.
If we don't have the value set we can safely operate on the local pte value
and we atomicaly set _PAGE_BUSY.
4) THP collapse:
A normal page gets converted to hugepage. In the collapse path, we
mark the pmd none early (pmdp_clear_flush). With irq disabled, if we
are aleady walking page table we would see the pmd_none and won't continue.
If we see a valid PMD, we should still check for _PAGE_PRESENT before
setting _PAGE_BUSY, to make sure we didn't collapse the PTE to a Huge PTE.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The deposted PTE page in the second half of the PMD table is used to
track the state on hash PTEs. After updating the HPTE, we mark the
coresponding slot in the deposted PTE page valid.
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Michael Ellerman