Binutils version up to (and including) 2.25 have a pathological
behaviour when it comes to mixing .inst directive and arithmetic
involving labels. The assembler complains about non-constant
expressions and compilation stops pretty quickly.

In order to detect this and work around it, let's add a bit of
detection code that will set the CONFIG_BROKEN_GAS_INST option
should a broken gas be detected.

Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The asm/opcodes.h file is now gone, but probes.h still references it
for not obvious reason. Removing the #include directive fixes
the compilation.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The opcodes.h drags in a lot of definition from the 32bit port, most
of which is not required at all. Clean things up a bit by moving
the bare minimum of what is required next to the actual users,
and drop the include file.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

Under CONFIG_DEBUG_PREEMPT=y, this_cpu_ptr() ends up calling back into
raw_smp_processor_id(), resulting in some hilariously catastrophic
infinite recursion. In the normal case, we have:

  #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr)

and everything is dandy. However for CONFIG_DEBUG_PREEMPT, this_cpu_ptr()
is defined in terms of my_cpu_offset, wherein the fun begins:

  #define my_cpu_offset per_cpu_offset(smp_processor_id())
  ...
  #define smp_processor_id() debug_smp_processor_id()
  ...
  notrace unsigned int debug_smp_processor_id(void)
  {
  	return check_preemption_disabled("smp_processor_id", "");
  ...
  notrace static unsigned int check_preemption_disabled(const char *what1,
  							const char *what2)
  {
  	int this_cpu = raw_smp_processor_id();

and bang. Use raw_cpu_ptr() directly to avoid that.

Fixes: 57c82954

 ("arm64: make cpu number a percpu variable")
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

* will/for-next/perf:
  selftests: arm64: add test for unaligned/inexact watchpoint handling
  arm64: Allow hw watchpoint of length 3,5,6 and 7
  arm64: hw_breakpoint: Handle inexact watchpoint addresses
  arm64: Allow hw watchpoint at varied offset from base address
  hw_breakpoint: Allow watchpoint of length 3,5,6 and 7

Bit positions of CNTHCTL_EL2 are changing depending on HCR_EL2.E2H bit.
EL1PCEN and EL1PCTEN are 1st and 0th bits when E2H is not set, but they
are 11th and 10th bits respectively when E2H is set.  Current code is
unintentionally setting wrong bits to CNTHCTL_EL2 with E2H set.

In fact, we don't need to set those two bits, which allow EL1 and EL0 to
access physical timer and counter respectively, if E2H and TGE are set
for the host kernel. They will be configured later as necessary. First,
we don't need to configure those bits for EL1, since the host kernel
runs in EL2.  It is a hypervisor's responsibility to configure them
before entering a VM, which runs in EL0 and EL1. Second, EL0 accesses
are configured in the later stage of boot process.

Signed-off-by: Jintack Lim <jintack@cs.columbia.edu>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The flush_cache_range() function (similarly for flush_cache_page()) is
called when the kernel is changing an existing VA->PA mapping range to
either a new PA or to different attributes. Since ARMv8 has PIPT-like
D-caches, this function does not need to perform any D-cache
maintenance. The I-cache maintenance is already handled via set_pte_at()
and flush_cache_range() cannot anyway guarantee that there are no cache
lines left after invalidation due to the speculative loads.

This patch makes flush_cache_range() a no-op.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch adds CONFIG_HIBERNATION to the arm64 defconfig.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch adds the Kconfig option to enable support for TTBR0 PAN
emulation. The option is default off because of a slight performance hit
when enabled, caused by the additional TTBR0_EL1 switching during user
access operations or exception entry/exit code.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

Privcmd calls are issued by the userspace. The kernel needs to enable
access to TTBR0_EL1 as the hypervisor would issue stage 1 translations
to user memory via AT instructions. Since AT instructions are not
affected by the PAN bit (ARMv8.1), we only need the explicit
uaccess_enable/disable if the TTBR0 PAN option is enabled.

Reviewed-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

When TTBR0_EL1 is set to the reserved page, an erroneous kernel access
to user space would generate a translation fault. This patch adds the
checks for the software-set PSR_PAN_BIT to emulate a permission fault
and report it accordingly.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

When the TTBR0 PAN feature is enabled, the kernel entry points need to
disable access to TTBR0_EL1. The PAN status of the interrupted context
is stored as part of the saved pstate, reusing the PSR_PAN_BIT (22).
Restoring access to TTBR0_EL1 is done on exception return if returning
to user or returning to a context where PAN was disabled.

Context switching via switch_mm() must defer the update of TTBR0_EL1
until a return to user or an explicit uaccess_enable() call.

Special care needs to be taken for two cases where TTBR0_EL1 is set
outside the normal kernel context switch operation: EFI run-time
services (via efi_set_pgd) and CPU suspend (via cpu_(un)install_idmap).
Code has been added to avoid deferred TTBR0_EL1 switching as in
switch_mm() and restore the reserved TTBR0_EL1 when uninstalling the
special TTBR0_EL1.

User cache maintenance (user_cache_maint_handler and
__flush_cache_user_range) needs the TTBR0_EL1 re-instated since the
operations are performed by user virtual address.

This patch also removes a stale comment on the switch_mm() function.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch adds the uaccess macros/functions to disable access to user
space by setting TTBR0_EL1 to a reserved zeroed page. Since the value
written to TTBR0_EL1 must be a physical address, for simplicity this
patch introduces a reserved_ttbr0 page at a constant offset from
swapper_pg_dir. The uaccess_disable code uses the ttbr1_el1 value
adjusted by the reserved_ttbr0 offset.

Enabling access to user is done by restoring TTBR0_EL1 with the value
from the struct thread_info ttbr0 variable. Interrupts must be disabled
during the uaccess_ttbr0_enable code to ensure the atomicity of the
thread_info.ttbr0 read and TTBR0_EL1 write. This patch also moves the
get_thread_info asm macro from entry.S to assembler.h for reuse in the
uaccess_ttbr0_* macros.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch takes the errata workaround code out of cpu_do_switch_mm into
a dedicated post_ttbr0_update_workaround macro which will be reused in a
subsequent patch.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch moves the directly coded alternatives for turning PAN on/off
into separate uaccess_{enable,disable} macros or functions. The asm
macros take a few arguments which will be used in subsequent patches.

Note that any (unlikely) access that the compiler might generate between
uaccess_enable() and uaccess_disable(), other than those explicitly
specified by the user access code, will not be protected by PAN.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch updates the description of the synchronous external aborts on
translation table walks.

Cc: Will Deacon <will.deacon@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

ARM64 hardware expects 64bit aligned address for watchpoint invocation.
However, it provides byte selection method to select any number of
consecutive byte set within the range of 1-8.

This patch adds support to test all such byte selection option for
different memory write sizes.

Patch also adds a test for handling the case when the cpu does not
report an address which exactly matches one of the regions we have
been watching (which is a situation permitted by the spec if an
instruction accesses both watched and unwatched regions). The test
was failing on a MSM8996pro before this patch series and is
passing now.

Signed-off-by: Pavel Labath <labath@google.com>
Signed-off-by: Pratyush Anand <panand@redhat.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

Since, arm64 can support all offset within a double word limit. Therefore,
now support other lengths within that range as well.

Signed-off-by: Pratyush Anand <panand@redhat.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

Arm64 hardware does not always report a watchpoint hit address that
matches one of the watchpoints set. It can also report an address
"near" the watchpoint if a single instruction access both watched and
unwatched addresses. There is no straight-forward way, short of
disassembling the offending instruction, to map that address back to
the watchpoint.

Previously, when the hardware reported a watchpoint hit on an address
that did not match our watchpoint (this happens in case of instructions
which access large chunks of memory such as "stp") the process would
enter a loop where we would be continually resuming it (because we did
not recognise that watchpoint hit) and it would keep hitting the
watchpoint again and again. The tracing process would never get
notified of the watchpoint hit.

This commit fixes the problem by looking at the watchpoints near the
address reported by the hardware. If the address does not exactly match
one of the watchpoints we have set, it attributes the hit to the
nearest watchpoint we have.  This heuristic is a bit dodgy, but I don't
think we can do much more, given the hardware limitations.

Signed-off-by: Pavel Labath <labath@google.com>
[panand: reworked to rebase on his patches]
Signed-off-by: Pratyush Anand <panand@redhat.com>
[will: use __ffs instead of ffs - 1]
Signed-off-by: Will Deacon <will.deacon@arm.com>

ARM64 hardware supports watchpoint at any double word aligned address.
However, it can select any consecutive bytes from offset 0 to 7 from that
base address. For example, if base address is programmed as 0x420030 and
byte select is 0x1C, then access of 0x420032,0x420033 and 0x420034 will
generate a watchpoint exception.

Currently, we do not have such modularity. We can only program byte,
halfword, word and double word access exception from any base address.

This patch adds support to overcome above limitations.

Signed-off-by: Pratyush Anand <panand@redhat.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

We only support breakpoint/watchpoint of length 1, 2, 4 and 8. If we can
support other length as well, then user may watch more data with less
number of watchpoints (provided hardware supports it). For example: if we
have to watch only 4th, 5th and 6th byte from a 64 bit aligned address, we
will have to use two slots to implement it currently. One slot will watch a
half word at offset 4 and other a byte at offset 6. If we can have a
watchpoint of length 3 then we can watch it with single slot as well.

ARM64 hardware does support such functionality, therefore adding these new
definitions in generic layer.

Signed-off-by: Pratyush Anand <panand@redhat.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

The arm64 kernel assumes that FP/ASIMD units are always present
and accesses the FP/ASIMD specific registers unconditionally. This
could cause problems when they are absent. This patch adds the
support for kernel handling systems without FP/ASIMD by skipping the
register access within the kernel. For kvm, we trap the accesses
to FP/ASIMD and inject an undefined instruction exception to the VM.

The callers of the exported kernel_neon_begin_partial() should
make sure that the FP/ASIMD is supported.

Cc: Will Deacon <will.deacon@arm.com>
Cc: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
[catalin.marinas@arm.com: add comment on the ARM64_HAS_NO_FPSIMD conflict and the new location]
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The hypervisor may not have full access to the kernel data structures
and hence cannot safely use cpus_have_cap() helper for checking the
system capability. Add a safe helper for hypervisors to check a constant
system capability, which *doesn't* fall back to checking the bitmap
maintained by the kernel. With this, make the cpus_have_cap() only
check the bitmask and force constant cap checks to use the new API
for quicker checks.

Cc: Robert Ritcher <rritcher@cavium.com>
Cc: Tirumalesh Chalamarla <tchalamarla@cavium.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

This patch moves arm64's struct thread_info from the task stack into
task_struct. This protects thread_info from corruption in the case of
stack overflows, and makes its address harder to determine if stack
addresses are leaked, making a number of attacks more difficult. Precise
detection and handling of overflow is left for subsequent patches.

Largely, this involves changing code to store the task_struct in sp_el0,
and acquire the thread_info from the task struct. Core code now
implements current_thread_info(), and as noted in <linux/sched.h> this
relies on offsetof(task_struct, thread_info) == 0, enforced by core
code.

This change means that the 'tsk' register used in entry.S now points to
a task_struct, rather than a thread_info as it used to. To make this
clear, the TI_* field offsets are renamed to TSK_TI_*, with asm-offsets
appropriately updated to account for the structural change.

Userspace clobbers sp_el0, and we can no longer restore this from the
stack. Instead, the current task is cached in a per-cpu variable that we
can safely access from early assembly as interrupts are disabled (and we
are thus not preemptible).

Both secondary entry and idle are updated to stash the sp and task
pointer separately.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: James Morse <james.morse@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

Shortly we will want to load a percpu variable in the return from
userspace path. We can save an instruction by folding the addition of
the percpu offset into the load instruction, and this patch adds a new
helper to do so.

At the same time, we clean up this_cpu_ptr for consistency. As with
{adr,ldr,str}_l, we change the template to take the destination register
first, and name this dst. Secondly, we rename the macro to adr_this_cpu,
following the scheme of adr_l, and matching the newly added
ldr_this_cpu.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: James Morse <james.morse@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

In the absence of CONFIG_THREAD_INFO_IN_TASK, core code maintains
thread_info::cpu, and low-level architecture code can access this to
build raw_smp_processor_id(). With CONFIG_THREAD_INFO_IN_TASK, core code
maintains task_struct::cpu, which for reasons of hte header soup is not
accessible to low-level arch code.

Instead, we can maintain a percpu variable containing the cpu number.

For both the old and new implementation of raw_smp_processor_id(), we
read a syreg into a GPR, add an offset, and load the result. As the
offset is now larger, it may not be folded into the load, but otherwise
the assembly shouldn't change much.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: James Morse <james.morse@arm.com>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

Subsequent patches will make smp_processor_id() use a percpu variable.
This will make smp_processor_id() dependent on the percpu offset, and
thus we cannot use smp_processor_id() to figure out what to initialise
the offset to.

Prepare for this by initialising the percpu offset based on
current::cpu, which will work regardless of how smp_processor_id() is
implemented. Also, make this relationship obvious by placing this code
together at the start of secondary_start_kernel().

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

When returning from idle, we rely on the fact that thread_info lives at
the end of the kernel stack, and restore this by masking the saved stack
pointer. Subsequent patches will sever the relationship between the
stack and thread_info, and to cater for this we must save/restore sp_el0
explicitly, storing it in cpu_suspend_ctx.

As cpu_suspend_ctx must be doubleword aligned, this leaves us with an
extra slot in cpu_suspend_ctx. We can use this to save/restore tpidr_el1
in the same way, which simplifies the code, avoiding pointer chasing on
the restore path (as we no longer need to load thread_info::cpu followed
by the relevant slot in __per_cpu_offset based on this).

This patch stashes both registers in cpu_suspend_ctx.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: James Morse <james.morse@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

When CONFIG_THREAD_INFO_IN_TASK is selected, task stacks may be freed
before a task is destroyed. To account for this, the stacks are
refcounted, and when manipulating the stack of another task, it is
necessary to get/put the stack to ensure it isn't freed and/or re-used
while we do so.

This patch reworks the arm64 stack walking code to account for this.
When CONFIG_THREAD_INFO_IN_TASK is not selected these perform no
refcounting, and this should only be a structural change that does not
affect behaviour.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The walk_stackframe functions is architecture-specific, with a varying
prototype, and common code should not use it directly. None of its
current users can be built as modules. With THREAD_INFO_IN_TASK, users
will also need to hold a stack reference before calling it.

There's no reason for it to be exported, and it's very easy to misuse,
so unexport it for now.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

In arm64's die and __die routines we pass around a thread_info, and
subsequently use this to determine the relevant task_struct, and the end
of the thread's stack. Subsequent patches will decouple thread_info from
the stack, and this approach will no longer work.

To figure out the end of the stack, we can use the new generic
end_of_stack() helper. As we only call __die() from die(), and die()
always deals with the current task, we can remove the parameter and have
both acquire current directly, which also makes it clear that __die
can't be called for arbitrary tasks.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

We define current_stack_pointer in <asm/thread_info.h>, though other
files and header relying upon it do not have this necessary include, and
are thus fragile to changes in the header soup.

Subsequent patches will affect the header soup such that directly
including <asm/thread_info.h> may result in a circular header include in
some of these cases, so we can't simply include <asm/thread_info.h>.

Instead, factor current_thread_info into its own header, and have all
existing users include this explicitly.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

Subsequent patches will move the thread_info::{task,cpu} fields, and the
current TI_{TASK,CPU} offset definitions are not used anywhere.

This patch removes the redundant definitions.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: James Morse <james.morse@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

We have a comment claiming __switch_to() cares about where cpu_context
is located relative to cpu_domain in thread_info. However arm64 has
never had a thread_info::cpu_domain field, and neither __switch_to nor
cpu_switch_to care where the cpu_context field is relative to others.

Additionally, the init_thread_info alias is never used anywhere in the
kernel, and will shortly become problematic when thread_info is moved
into task_struct.

This patch removes both.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: James Morse <james.morse@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

When CONFIG_THREAD_INFO_IN_TASK is selected, the current_thread_info()
macro relies on current having been defined prior to its use. However,
not all users of current_thread_info() include <asm/current.h>, and thus
current is not guaranteed to be defined.

When CONFIG_THREAD_INFO_IN_TASK is not selected, it's possible that
get_current() / current are based upon current_thread_info(), and
<asm/current.h> includes <asm/thread_info.h>. Thus always including
<asm/current.h> would result in circular dependences on some platforms.

To ensure both cases work, this patch includes <asm/current.h>, but only
when CONFIG_THREAD_INFO_IN_TASK is selected.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: Andy Lutomirski <luto@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

Since commit f56141e3

 ("all arches, signal: move restart_block
to struct task_struct"), thread_info and restart_block have been
logically distinct, yet struct restart_block is still defined in
<linux/thread_info.h>.

At least one architecture (erroneously) uses restart_block as part of
its thread_info, and thus the definition of restart_block must come
before the include of <asm/thread_info>. Subsequent patches in this
series need to shuffle the order of includes and definitions in
<linux/thread_info.h>, and will make this ordering fragile.

This patch moves the definition of restart_block out to its own header.
This serves as generic cleanup, logically separating thread_info and
restart_block, and also makes it easier to avoid fragility.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Andy Lutomirski <luto@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

For several reasons it is preferable to use {READ,WRITE}_ONCE() rather than
ACCESS_ONCE(). For example, these handle aggregate types, result in shorter
source code, and better document the intended access (which may be useful for
instrumentation features such as the upcoming KTSAN).

Over a number of patches, most uses of ACCESS_ONCE() in arch/arm64 have been
migrated to {READ,WRITE}_ONCE(). For consistency, and the above reasons, this
patch migrates the final remaining uses.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The libhugetlbfs meets several failures since the following functions
do not use the correct address:
   huge_ptep_get_and_clear()
   huge_ptep_set_access_flags()
   huge_ptep_set_wrprotect()
   huge_ptep_clear_flush()

This patch fixes the wrong address for them.

Signed-off-by: Huang Shijie <shijie.huang@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The find_num_contig() will return 1 when the pmd is not present.
It will cause a kernel dead loop in the following scenaro:

   1.) pmd entry is not present.

   2.) the page fault occurs:
       ... hugetlb_fault() --> hugetlb_no_page() --> set_huge_pte_at()

   3.) set_huge_pte_at() will only set the first PMD entry, since the
       find_num_contig just return 1 in this case. So the PMD entries
       are all empty except the first one.

   4.) when kernel accesses the address mapped by the second PMD entry,
       a new page fault occurs:
       ... hugetlb_fault() --> huge_ptep_set_access_flags()

       The second PMD entry is still empty now.

   5.) When the kernel returns, the access will cause a page fault again.
       The kernel will run like the "4)" above.
       We will see a dead loop since here.

The dead loop is caught in the 32M hugetlb page (2M PMD + Contiguous bit).

This patch removes wrong pmd check, and fixes this dead loop.

This patch also removes the redundant checks for PGD/PUD in
the find_num_contig().

Acked-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Huang Shijie <shijie.huang@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

The default hugepage size when 64K pages are enabled is set to 2MB using
the contiguous PTE bit. The add_default_hugepagesz(), however, uses
CONT_PMD_SHIFT instead of CONT_PTE_SHIFT. There is no functional change
since the values are the same.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>