Our current switch_mm implementation suffers from a number of problems:
(1) The ASID allocator relies on IPIs to synchronise the CPUs on a
rollover event
(2) Because of (1), we cannot allocate ASIDs with interrupts disabled
and therefore make use of a TIF_SWITCH_MM flag to postpone the
actual switch to finish_arch_post_lock_switch
(3) We run context switch with a reserved (invalid) TTBR0 value, even
though the ASID and pgd are updated atomically
(4) We take a global spinlock (cpu_asid_lock) during context-switch
(5) We use h/w broadcast TLB operations when they are not required
(e.g. in flush_context)
This patch addresses these problems by rewriting the ASID algorithm to
match the bitmap-based arch/arm/ implementation more closely. This in
turn allows us to remove much of the complications surrounding switch_mm,
including the ugly thread flag.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
When cold-booting a CPU, we must invalidate any junk entries from the
local TLB prior to enabling the MMU. This doesn't require broadcasting
within the inner-shareable domain, so de-scope the operation to apply
only to the local CPU.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
We don't want to expose the DCC to userspace, particularly as there is
a kernel console driver for it.
This patch resets mdscr_el1 to disable userspace access to the DCC
registers on the cold boot path.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The arm64 booting document requires that the bootloader has cleaned the
kernel image to the PoC. However, when a CPU re-enters the kernel due to
either a CPU hotplug "on" event or resuming from a low-power state (e.g.
cpuidle), the kernel text may in-fact be dirty at the PoU due to things
like alternative patching or even module loading.
Thanks to I-cache speculation with the MMU off, stale instructions could
be fetched prior to enabling the MMU, potentially leading to crashes
when executing regions of code that have been modified at runtime.
This patch addresses the issue by ensuring that the local I-cache is
invalidated immediately after a CPU has enabled its MMU but before
jumping out of the identity mapping. Any stale instructions fetched from
the PoC will then be discarded and refetched correctly from the PoU.
Patching kernel text executed prior to the MMU being enabled is
prohibited, so the early entry code will always be clean.
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Nobody seems to be producing !SMP systems anymore, so this is just
becoming a source of kernel bugs, particularly if people want to use
coherent DMA with non-shared pages.
This patch forces CONFIG_SMP=y for arm64, removing a modest amount of
code in the process.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The ARMv8.1 architecture extensions introduce support for hardware
updates of the access and dirty information in page table entries. With
TCR_EL1.HA enabled, when the CPU accesses an address with the PTE_AF bit
cleared in the page table, instead of raising an access flag fault the
CPU sets the actual page table entry bit. To ensure that kernel
modifications to the page tables do not inadvertently revert a change
introduced by hardware updates, the exclusive monitor (ldxr/stxr) is
adopted in the pte accessors.
When TCR_EL1.HD is enabled, a write access to a memory location with the
DBM (Dirty Bit Management) bit set in the corresponding pte
automatically clears the read-only bit (AP[2]). Such DBM bit maps onto
the Linux PTE_WRITE bit and to check whether a writable (DBM set) page
is dirty, the kernel tests the PTE_RDONLY bit. In order to allow
read-only and dirty pages, the kernel needs to preserve the software
dirty bit. The hardware dirty status is transferred to the software
dirty bit in ptep_set_wrprotect() (using load/store exclusive loop) and
pte_modify().
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The documented semantics of flush_cache_all are not possible to provide
for arm64 (short of flushing the entire physical address space by VA),
and there are currently no users; KVM uses VA maintenance exclusively,
cpu_reset is never called, and the only two users outside of arch code
cannot be built for arm64.
While cpu_soft_reset and related functions (which call flush_cache_all)
were thought to be useful for kexec, their current implementations only
serve to mask bugs. For correctness kexec will need to perform
maintenance by VA anyway to account for system caches, line migration,
and other subtleties of the cache architecture. As the extent of this
cache maintenance will be kexec-specific, it should probably live in the
kexec code.
This patch removes flush_cache_all, and related unused components,
preventing further abuse.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: Geoff Levand <geoff@infradead.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The page size and the number of translation levels, and hence the supported
virtual address range, are build-time configurables on arm64 whose optimal
values are use case dependent. However, in the current implementation, if
the system's RAM is located at a very high offset, the virtual address range
needs to reflect that merely because the identity mapping, which is only used
to enable or disable the MMU, requires the extended virtual range to map the
physical memory at an equal virtual offset.
This patch relaxes that requirement, by increasing the number of translation
levels for the identity mapping only, and only when actually needed, i.e.,
when system RAM's offset is found to be out of reach at runtime.
Tested-by: Laura Abbott <lauraa@codeaurora.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Tested-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
ARM64_CPU_SUSPEND config option was introduced to make code providing
context save/restore selectable only on platforms requiring power
management capabilities.
Currently ARM64_CPU_SUSPEND depends on the PM_SLEEP config option which
in turn is set by the SUSPEND config option.
The introduction of CPU_IDLE for arm64 requires that code configured
by ARM64_CPU_SUSPEND (context save/restore) should be compiled in
in order to enable the CPU idle driver to rely on CPU operations
carrying out context save/restore.
The ARM64_CPUIDLE config option (ARM64 generic idle driver) is therefore
forced to select ARM64_CPU_SUSPEND, even if there may be (ie PM_SLEEP)
failed dependencies, which is not a clean way of handling the kernel
configuration option.
For these reasons, this patch removes the ARM64_CPU_SUSPEND config option
and makes the context save/restore dependent on CPU_PM, which is selected
whenever either SUSPEND or CPU_IDLE are configured, cleaning up dependencies
in the process.
This way, code previously configured through ARM64_CPU_SUSPEND is
compiled in whenever a power management subsystem requires it to be
present in the kernel (SUSPEND || CPU_IDLE), which is the behaviour
expected on ARM64 kernels.
The cpu_suspend and cpu_init_idle CPU operations are added only if
CPU_IDLE is selected, since they are CPU_IDLE specific methods and
should be grouped and defined accordingly.
PSCI CPU operations are updated to reflect the introduced changes.
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
We initialise the SCTLR_EL1 value by read-modify-writeback
of the desired bits, leaving the other bits (including reserved
bits(RESx)) untouched. However, sometimes the boot monitor could
leave garbage values in the RESx bits which could have different
implications. This patch makes sure that all the bits, including
the RESx bits, are set to the proper state, except for the
'endianness' control bits, EE(25) & E0E(24)- which are set early
in the el2_setup.
Updated the state of the Bit[6] in the comment to RES0 in the
comment.
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The current soft_restart() and setup_restart implementations incorrectly
assume that compiler will not spill/fill values to/from stack. However
this assumption seems to be wrong, revealed by the disassembly of the
currently existing code (v3.16) built with Linaro GCC 4.9-2014.05.
ffffffc000085224 <soft_restart>:
ffffffc000085224: a9be7bfd stp x29, x30, [sp,#-32]!
ffffffc000085228: 910003fd mov x29, sp
ffffffc00008522c: f9000fa0 str x0, [x29,#24]
ffffffc000085230: 94003d21 bl ffffffc0000946b4 <setup_mm_for_reboot>
ffffffc000085234: 94003b33 bl ffffffc000093f00 <flush_cache_all>
ffffffc000085238: 94003dfa bl ffffffc000094a20 <cpu_cache_off>
ffffffc00008523c: 94003b31 bl ffffffc000093f00 <flush_cache_all>
ffffffc000085240: b0003321 adrp x1, ffffffc0006ea000 <reset_devices>
ffffffc000085244: f9400fa0 ldr x0, [x29,#24] ----> spilled addr
ffffffc000085248: f942fc22 ldr x2, [x1,#1528] ----> global memstart_addr
ffffffc00008524c: f0000061 adrp x1, ffffffc000094000 <__inval_cache_range+0x40>
ffffffc000085250: 91290021 add x1, x1, #0xa40
ffffffc000085254: 8b010041 add x1, x2, x1
ffffffc000085258: d2c00802 mov x2, #0x4000000000 // #274877906944
ffffffc00008525c: 8b020021 add x1, x1, x2
ffffffc000085260: d63f0020 blr x1
...
Here the compiler generates memory accesses after the cache is disabled,
loading stale values for the spilled value and global variable. As we cannot
control when the compiler will access memory we must rewrite the
functions in assembly to stash values we need in registers prior to
disabling the cache, avoiding the use of memory.
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Arun Chandran <achandran@mvista.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In order to ensure ordering and completion of inner-shareable maintenance
instructions (cache and TLB) on AArch64, we can use the -ish suffix to
the dmb and dsb instructions respectively.
This patch updates our low-level cache and tlb maintenance routines to
use the inner-shareable barrier variants where appropriate.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The current TCR register setting in arch/arm64/mm/proc.S assumes that
TCR_EL1.TG* fields are one bit wide and bit 31 is RES1 (reserved, set to
1). With the addition of 16K pages (currently unsupported in the
kernel), the TCR_EL1.TG* fields have been extended to two bits. This
patch updates the corresponding Linux definitions and drops the bit 31
setting in proc.S in favour of the new macros.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Reported-by: Joe Sylve <joe.sylve@gmail.com>
ARMv8 supports a range of physical address bit sizes. The PARange bits
from ID_AA64MMFR0_EL1 register are read during boot-time and the
intermediate physical address size bits are written in the translation
control registers (TCR_EL1 and VTCR_EL2).
There is no change in the VA bits and levels of translation.
Signed-off-by: Radha Mohan Chintakuntla <rchintakuntla@cavium.com>
Reviewed-by: Will Deacon <Will.deacon@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Currently we flush the entire dcache at boot within __cpu_setup, but
this is unnecessary as the booting protocol demands that the dcache is
invalid and off upon entering the kernel. The presence of the cache
flush only serves to hide bugs in bootloaders, and is not safe in the
presence of SMP.
In an SMP boot scenario the CPUs enter coherency outside of the kernel,
and the primary CPU enables its caches before bringing up secondary
CPUs. Therefore if any secondary CPU has an entry in its cache (in
violation of the boot protocol), the primary CPU might snoop it even if
the secondary CPU's cache is disabled. The boot-time cache flush only
serves to hide a firmware bug, and slows down a cpu boot unnecessarily.
This patch removes the unnecessary boot-time cache flush.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
[catalin.marinas@arm.com: make __flush_dcache_all local only]
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Fix the function name of comment of cpu_do_switch_mm,
because cpu_do_switch_mm is the correct name.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Power management software requires the kernel to save and restore
CPU registers while going through suspend and resume operations
triggered by kernel subsystems like CPU idle and suspend to RAM.
This patch implements code that provides save and restore mechanism
for the arm v8 implementation. Memory for the context is passed as
parameter to both cpu_do_suspend and cpu_do_resume functions, and allows
the callers to implement context allocation as they deem fit.
The registers that are saved and restored correspond to the registers set
actually required by the kernel to be up and running which represents a
subset of v8 ISA.
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Currently there is no dsb between the tlbi in __cpu_setup and the write
to SCTLR_EL1 which enables the MMU in __turn_mmu_on. This means that the
TLB invalidation is not guaranteed to have completed at the point
address translation is enabled, leading to a number of possible issues
including incorrect translations and TLB conflict faults.
This patch moves the tlbi in __cpu_setup above an existing dsb used to
synchronise I-cache invalidation, ensuring that the TLBs have been
invalidated at the point the MMU is enabled.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The endianness of memory accesses at EL2 and EL1 are configured by
SCTLR_EL2.EE and SCTLR_EL1.EE respectively. When the kernel is booted,
the state of SCTLR_EL{2,1}.EE is unknown, and thus the kernel must
ensure that they are set before performing any memory accesses.
This patch ensures that SCTLR_EL{2,1} are configured appropriately at
boot for kernels of either endianness.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Matthew Leach <matthew.leach@arm.com>
[catalin.marinas@arm.com: fix SCTLR_EL1.E0E bit setting in head.S]
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
TCR.TBI0 can be used to cause hardware address translation to ignore the
top byte of userspace virtual addresses. Whilst not especially useful in
standard C programs, this can be used by JITs to `tag' pointers with
various pieces of metadata.
This patch enables this bit for AArch64 Linux, and adds a new file to
Documentation/arm64/ which describes some potential caveats when using
tagged virtual addresses.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
During boot, we take the debug OS lock before interrupts are enabled.
This is required to prevent clearing of PSTATE.D on the interrupt entry
path, which could result in spurious debug exceptions before we've got
round to resetting things like the hardware breakpoints registers to a
sane state.
A problem with this approach is that taking the OS lock prevents an
external JTAG debugger from debugging the system, which is especially
irritating during boot, where JTAG debugging can be most useful.
This patch clears mdscr_el1 rather than taking the lock, clearing the
MDE and KDE bits and preventing self-hosted hardware debug exceptions
from occurring.
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: stable@vger.kernel.org
If such bit exists on a given CPU, it must be set by the firmware or
boot-loader prior to starting the kernel (see
Documentation/arm64/booting.txt).
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
This patch adds AArch64 CPU specific functionality. It assumes that the
implementation is generic to AArch64 and does not require specific
identification. Different CPU implementations may require the setting of
various ACTLR_EL1 bits but such information is not currently available
and it should ideally be pushed to firmware.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>