platform/kernel/linux-rpi.git
7 years agoarm64: fix endianness annotation in get_kaslr_seed()
Luc Van Oostenryck [Thu, 29 Jun 2017 14:35:29 +0000 (16:35 +0200)]
arm64: fix endianness annotation in get_kaslr_seed()

In the flattened device tree format, all integer properties are
in big-endian order.
Here the property "kaslr-seed" is read from the fdt and then
correctly converted to native order (via fdt64_to_cpu()) but the
pointer used for this is not annotated as being for big-endian.

Fix this by declaring the pointer as fdt64_t instead of u64
(fdt64_t being itself typedefed to __be64).

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: add missing conversion to __wsum in ip_fast_csum()
Luc Van Oostenryck [Thu, 29 Jun 2017 14:31:40 +0000 (16:31 +0200)]
arm64: add missing conversion to __wsum in ip_fast_csum()

ARM64 implementation of ip_fast_csum() do most of the work
in 128 or 64 bit and call csum_fold() to finalize. csum_fold()
itself take a __wsum argument, to insure that this value is
always a 32bit native-order value.

Fix this by adding the sadly needed '__force' to cast the native
'sum' to the type '__wsum'.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fix endianness annotation in acpi_parking_protocol.c
Luc Van Oostenryck [Wed, 28 Jun 2017 14:58:07 +0000 (16:58 +0200)]
arm64: fix endianness annotation in acpi_parking_protocol.c

Here both variables 'cpu_id' and 'entry_point' are read via
read[lq]_relaxed(), from a little-endian annotated pointer
and then used as a native endian value.

This is correct since the read[lq]() family of function
internally do a little-to-native endian conversion.

But in this case, it is wrong to declare these variable as
little-endian since there are native ones.

Fix this by changing the declaration of these variables
as 'u32' or 'u64' instead of '__le32' / '__le64'.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: use readq() instead of readl() to read 64bit entry_point
Luc Van Oostenryck [Mon, 26 Jun 2017 13:16:25 +0000 (15:16 +0200)]
arm64: use readq() instead of readl() to read 64bit entry_point

Here the entrypoint, declared as a 64 bit integer, is read from
a pointer to 64bit integer but the read is done via readl_relaxed()
which is for 32bit quantities.

All the high bits will thus be lost which change the meaning
of the test against zero done later.

Fix this by using readq_relaxed() instead as it should be for
64bit quantities.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fix endianness annotation for reloc_insn_movw() & reloc_insn_imm()
Luc Van Oostenryck [Wed, 28 Jun 2017 14:56:00 +0000 (16:56 +0200)]
arm64: fix endianness annotation for reloc_insn_movw() & reloc_insn_imm()

Here the functions reloc_insn_movw() & reloc_insn_imm() are used
to read, modify and write back ARM instructions, which are always
stored in memory in little-endian order. These values are thus
correctly converted to/from native order but the pointers used to
hold their addresses are declared as for native order values.

Fix this by declaring the pointers as __le32* and remove the
casts that are now unneeded.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fix endianness annotation for aarch64_insn_write()
Luc Van Oostenryck [Wed, 28 Jun 2017 14:58:11 +0000 (16:58 +0200)]
arm64: fix endianness annotation for aarch64_insn_write()

aarch64_insn_write() is used to write an instruction.
As on ARM64 in-memory instructions are always stored
in little-endian order, this function, taking the instruction
opcode in native order, correctly convert it to little-endian
before sending it to an helper function __aarch64_insn_write()
which will do the effective write.

This is all good, but the variable and argument holding the
converted value are not annotated for a little-endian value
but left for native values.

Fix this by adjusting the prototype of the helper and
directly using the result of cpu_to_le32() without passing
by an intermediate variable (which was not a distinct one
but the same as the one holding the native value).

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fix endianness annotation in aarch64_insn_read()
Luc Van Oostenryck [Wed, 28 Jun 2017 14:58:09 +0000 (16:58 +0200)]
arm64: fix endianness annotation in aarch64_insn_read()

The function arch64_insn_read() is used to read an instruction.
On AM64 instructions are always stored in little-endian order
and thus the function correctly do a little-to-native endian
conversion to the value just read.

However, the variable used to hold the value before the conversion
is not declared for a little-endian value but for a native one.

Fix this by using the correct type for the declaration: __le32

Note: This only works because the function reading the value,
      probe_kernel_read((), takes a void pointer and void pointers
      are endian-agnostic. Otherwise probe_kernel_read() should
      also be properly annotated (or worse, need to be specialized).

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fix endianness annotation in call_undef_hook()
Luc Van Oostenryck [Wed, 28 Jun 2017 14:55:55 +0000 (16:55 +0200)]
arm64: fix endianness annotation in call_undef_hook()

Here we're reading thumb or ARM instructions, which are always
stored in memory in little-endian order. These values are thus
correctly converted to native order but the intermediate value
should be annotated as for little-endian values.

Fix this by declaring the intermediate var as __le32 or __le16.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fix endianness annotation for debug-monitors.c
Luc Van Oostenryck [Wed, 28 Jun 2017 14:55:52 +0000 (16:55 +0200)]
arm64: fix endianness annotation for debug-monitors.c

Here we're reading thumb or ARM instructions, which are always
stored in memory in little-endian order. These values are thus
correctly converted to native order but the intermediate value
should be annotated as for little-endian values.

Fix this by declaring the intermediate var as __le32 or __le16.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoras: mark stub functions as 'inline'
Arnd Bergmann [Tue, 27 Jun 2017 15:35:41 +0000 (17:35 +0200)]
ras: mark stub functions as 'inline'

With CONFIG_RAS disabled, we get two harmless warnings about
unused functions:

include/linux/ras.h:37:13: error: 'log_arm_hw_error' defined but not used [-Werror=unused-function]
 static void log_arm_hw_error(struct cper_sec_proc_arm *err) { return; }
include/linux/ras.h:33:13: error: 'log_non_standard_event' defined but not used [-Werror=unused-function]
 static void log_non_standard_event(const guid_t *sec_type,

Clearly these are meant to be 'inline', like the other stubs
in the same header.

Fixes: 297b64c74385 ("ras: acpi / apei: generate trace event for unrecognized CPER section")
Fixes: e9279e83ad1f ("trace, ras: add ARM processor error trace event")
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoMerge branch 'aarch64/for-next/ras-apei' into aarch64/for-next/core
Will Deacon [Mon, 26 Jun 2017 09:54:27 +0000 (10:54 +0100)]
Merge branch 'aarch64/for-next/ras-apei' into aarch64/for-next/core

Merge in arm64 ACPI RAS support (APEI/GHES) from Tyler Baicar.

7 years agoMerge branch 'perf/updates' into aarch64/for-next/core
Will Deacon [Mon, 26 Jun 2017 09:50:50 +0000 (10:50 +0100)]
Merge branch 'perf/updates' into aarch64/for-next/core

Merge in arm64 perf updates:

  * xgene system PMUv3 support
  * 16-bit events for ARMv8.1

7 years agoarm64: pass endianness info to sparse
Luc Van Oostenryck [Sat, 24 Jun 2017 15:42:11 +0000 (17:42 +0200)]
arm64: pass endianness info to sparse

ARM64 depends on the macro __AARCH64EB__ being defined or not
to correctly select or define endian-specific macros, structures
or pieces of code.

This macro is predefined by the compiler but sparse knows nothing
about it and thus may pre-process files differently from what
gcc would.

Fix this by passing '-D__AARCH64EL__' or '-D__AARCH64EB__' to
sparse depending of the endianness of the kernel, like defined
by GCC.

Note: In most case it won't change anything since most arm64 use
      little-endian (but an allyesconfig would use big-endian!).

CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: linux-arm-kernel@lists.infradead.org
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: ftrace: fix !CONFIG_ARM64_MODULE_PLTS kernels
Mark Rutland [Fri, 23 Jun 2017 17:02:06 +0000 (18:02 +0100)]
arm64: ftrace: fix !CONFIG_ARM64_MODULE_PLTS kernels

When a kernel is built without CONFIG_ARM64_MODULE_PLTS, we don't
generate the expected branch instruction in ftrace_make_nop(). This
means we pass zero (rather than a valid branch) to ftrace_modify_code()
as the expected instruction to validate. This causes us to return
-EINVAL to the core ftrace code for a valid case, resulting in a splat
at boot time.

This was an unintended effect of commit:

  687644209a6e9557 ("arm64: ftrace: fix building without CONFIG_MODULES")

... which incorrectly moved the generation of the branch instruction
into the ifdef for CONFIG_ARM64_MODULE_PLTS.

This patch fixes the issue by moving the ifdef inside of the relevant
if-else case, and always checking that the branch is in range,
regardless of CONFIG_ARM64_MODULE_PLTS. This ensures that we generate
the expected branch instruction, and also improves our sanity checks.

For consistency, both ftrace_make_nop() and ftrace_make_call() are
updated with this pattern.

Fixes: 687644209a6e9557 ("arm64: ftrace: fix building without CONFIG_MODULES")
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reported-by: Marc Zyngier <marc.zyngier@arm.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: signal: Allow expansion of the signal frame
Dave Martin [Tue, 20 Jun 2017 17:23:39 +0000 (18:23 +0100)]
arm64: signal: Allow expansion of the signal frame

This patch defines an extra_context signal frame record that can be
used to describe an expanded signal frame, and modifies the context
block allocator and signal frame setup and parsing code to create,
populate, parse and decode this block as necessary.

To avoid abuse by userspace, parse_user_sigframe() attempts to
ensure that:

 * no more than one extra_context is accepted;
 * the extra context data is a sensible size, and properly placed
   and aligned.

The extra_context data is required to start at the first 16-byte
aligned address immediately after the dummy terminator record
following extra_context in rt_sigframe.__reserved[] (as ensured
during signal delivery).  This serves as a sanity-check that the
signal frame has not been moved or copied without taking the extra
data into account.

Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
[will: add __force annotation when casting extra_datap to __user pointer]
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoacpi: apei: check for pending errors when probing GHES entries
Tyler Baicar [Wed, 21 Jun 2017 18:17:15 +0000 (12:17 -0600)]
acpi: apei: check for pending errors when probing GHES entries

Check for pending errors when probing GHES entries. It is possible
that a fatal error is already pending at this point, so we should
handle it as soon as the driver is probed. This also avoids a
potential issue if there was an interrupt that was already
cleared for an error since the GHES driver wasn't present.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoperf: xgene: Add support for SoC PMU version 3
Hoan Tran [Thu, 22 Jun 2017 18:26:05 +0000 (19:26 +0100)]
perf: xgene: Add support for SoC PMU version 3

This patch adds support for SoC-wide (AKA uncore) Performance Monitoring
Unit version 3.

It can support up to
 - 2 IOB PMU instances
 - 8 L3C PMU instances
 - 2 MCB PMU instances
 - 8 MCU PMU instances
and these PMUs support 64 bit counter

Signed-off-by: Hoan Tran <hotran@apm.com>
[Mark: stop counters in _xgene_pmu_isr()]
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
[will: make xgene_pmu_v3_ops static]
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoperf: xgene: Move PMU leaf functions into function pointer structure
Hoan Tran [Thu, 22 Jun 2017 18:26:04 +0000 (19:26 +0100)]
perf: xgene: Move PMU leaf functions into function pointer structure

This patch moves PMU leaf functions into a function pointer structure.
It helps code maintain and expasion easier.

Signed-off-by: Hoan Tran <hotran@apm.com>
[Mark: remove redundant cast]
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
[will: make xgene_pmu_ops static]
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoperf: xgene: Parse PMU subnode from the match table
Hoan Tran [Thu, 22 Jun 2017 18:26:03 +0000 (19:26 +0100)]
perf: xgene: Parse PMU subnode from the match table

This patch parses PMU Subnode from a match table.

Signed-off-by: Hoan Tran <hotran@apm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm/arm64: KVM: add guest SEA support
Tyler Baicar [Wed, 21 Jun 2017 18:17:14 +0000 (12:17 -0600)]
arm/arm64: KVM: add guest SEA support

Currently external aborts are unsupported by the guest abort
handling. Add handling for SEAs so that the host kernel reports
SEAs which occur in the guest kernel.

When an SEA occurs in the guest kernel, the guest exits and is
routed to kvm_handle_guest_abort(). Prior to this patch, a print
message of an unsupported FSC would be printed and nothing else
would happen. With this patch, the code gets routed to the APEI
handling of SEAs in the host kernel to report the SEA information.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Christoffer Dall <cdall@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agotrace, ras: add ARM processor error trace event
Tyler Baicar [Wed, 21 Jun 2017 18:17:13 +0000 (12:17 -0600)]
trace, ras: add ARM processor error trace event

Currently there are trace events for the various RAS
errors with the exception of ARM processor type errors.
Add a new trace event for such errors so that the user
will know when they occur. These trace events are
consistent with the ARM processor error section type
defined in UEFI 2.6 spec section N.2.4.4.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Reviewed-by: Xie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoras: acpi / apei: generate trace event for unrecognized CPER section
Tyler Baicar [Wed, 21 Jun 2017 18:17:12 +0000 (12:17 -0600)]
ras: acpi / apei: generate trace event for unrecognized CPER section

The UEFI spec includes non-standard section type support in the
Common Platform Error Record. This is defined in section N.2.3 of
UEFI version 2.5.

Currently if the CPER section's type (UUID) does not match any
section type that the kernel knows how to parse, a trace event is
not generated.

Generate a trace event which contains the raw error data for
non-standard section type error records.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Tested-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoefi: print unrecognized CPER section
Tyler Baicar [Wed, 21 Jun 2017 18:17:11 +0000 (12:17 -0600)]
efi: print unrecognized CPER section

UEFI spec allows for non-standard section in Common Platform Error
Record. This is defined in section N.2.3 of UEFI version 2.5.

Currently if the CPER section's type (UUID) does not match with
one of the section types that the kernel knows how to parse, the
section is skipped. Therefore, user is not able to see
such CPER data, for instance, error record of non-standard section.

This change prints out the raw data in hex in the dmesg buffer so
that non-standard sections are reported to the user. Non-standard
section type errors should be reported to the user because these
can include errors which are vendor specific. The data length is
taken from Error Data length field of Generic Error Data Entry.

The following is a sample output from dmesg:
 Hardware error from APEI Generic Hardware Error Source: 2
 It has been corrected by h/w and requires no further action
 event severity: corrected
  time: precise 2017-03-15 20:37:35
  Error 0, type: corrected
   section type: unknown, d2e2621c-f936-468d-0d84-15a4ed015c8b
   section length: 0x238
   000000004d415201 4d492031 453a4d45 435f4343  .RAM1 IMEM:ECC_C
   0000001053515f45 44525f42 00000000 00000000  E_QSB_RD........
   0000002000000000 00000000 00000000 00000000  ................
   0000003000000000 00000000 01010000 01010000  ................
   0000004000000000 00000000 00000005 00000000  ................
   0000005001010000 00000000 00000001 00dddd00  ................
...

The raw data from the error can then be decoded using vendor
specific tools.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Reviewed-by: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoacpi: apei: panic OS with fatal error status block
Jonathan (Zhixiong) Zhang [Wed, 21 Jun 2017 18:17:10 +0000 (12:17 -0600)]
acpi: apei: panic OS with fatal error status block

Even if an error status block's severity is fatal, the kernel does not
honor the severity level and panic.

With the firmware first model, the platform could inform the OS about a
fatal hardware error through the non-NMI GHES notification type. The OS
should panic when a hardware error record is received with this
severity.

Call panic() after CPER data in error status block is printed if
severity is fatal, before each error section is handled.

Signed-off-by: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
Reviewed-by: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoacpi: apei: handle SEA notification type for ARMv8
Tyler Baicar [Wed, 21 Jun 2017 18:17:09 +0000 (12:17 -0600)]
acpi: apei: handle SEA notification type for ARMv8

ARM APEI extension proposal added SEA (Synchronous External Abort)
notification type for ARMv8.
Add a new GHES error source handling function for SEA. If an error
source's notification type is SEA, then this function can be registered
into the SEA exception handler. That way GHES will parse and report
SEA exceptions when they occur.
An SEA can interrupt code that had interrupts masked and is treated as
an NMI. To aid this the page of address space for mapping APEI buffers
while in_nmi() is always reserved, and ghes_ioremap_pfn_nmi() is
changed to use the helper methods to find the prot_t to map with in
the same way as ghes_ioremap_pfn_irq().

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Reviewed-by: James Morse <james.morse@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: exception: handle Synchronous External Abort
Tyler Baicar [Wed, 21 Jun 2017 18:17:08 +0000 (12:17 -0600)]
arm64: exception: handle Synchronous External Abort

SEA exceptions are often caused by an uncorrected hardware
error, and are handled when data abort and instruction abort
exception classes have specific values for their Fault Status
Code.
When SEA occurs, before killing the process, report the error
in the kernel logs.
Update fault_info[] with specific SEA faults so that the
new SEA handler is used.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Reviewed-by: James Morse <james.morse@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
[will: use NULL instead of 0 when assigning si_addr]
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: dump cpu_hwcaps at panic time
Mark Rutland [Wed, 21 Jun 2017 17:11:23 +0000 (18:11 +0100)]
arm64: dump cpu_hwcaps at panic time

When debugging a kernel panic(), it can be useful to know which CPU
features have been detected by the kernel, as some code paths can depend
on these (and may have been patched at runtime).

This patch adds a notifier to dump the detected CPU caps (as a hex
string) at panic(), when we log other information useful for debugging.
On a Juno R1 system running v4.12-rc5, this looks like:

[  615.431249] Kernel panic - not syncing: Fatal exception in interrupt
[  615.437609] SMP: stopping secondary CPUs
[  615.441872] Kernel Offset: disabled
[  615.445372] CPU features: 0x02086
[  615.448522] Memory Limit: none

A developer can decode this by looking at the corresponding
<asm/cpucaps.h> bits. For example, the above decodes as:

* bit  1: ARM64_WORKAROUND_DEVICE_LOAD_ACQUIRE
* bit  2: ARM64_WORKAROUND_845719
* bit  7: ARM64_WORKAROUND_834220
* bit 13: ARM64_HAS_32BIT_EL0

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Steve Capper <steve.capper@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: ptrace: Flush user-RW TLS reg to thread_struct before reading
Dave Martin [Wed, 21 Jun 2017 15:00:44 +0000 (16:00 +0100)]
arm64: ptrace: Flush user-RW TLS reg to thread_struct before reading

When reading current's user-writable TLS register (which occurs
when dumping core for native tasks), it is possible that userspace
has modified it since the time the task was last scheduled out.
The new TLS register value is not guaranteed to have been written
immediately back to thread_struct in this case.

As a result, a coredump can capture stale data for this register.
Reading the register for a stopped task via ptrace is unaffected.

For native tasks, this patch explicitly flushes the TPIDR_EL0
register back to thread_struct before dumping when operating on
current, thus ensuring that coredump contents are up to date.  For
compat tasks, the TLS register is not user-writable and so cannot
be out of sync, so no flush is required in compat_tls_get().

Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: ptrace: Flush FPSIMD regs back to thread_struct before reading
Dave Martin [Wed, 21 Jun 2017 15:00:43 +0000 (16:00 +0100)]
arm64: ptrace: Flush FPSIMD regs back to thread_struct before reading

When reading the FPSIMD state of current (which occurs when dumping
core), it is possible that userspace has modified the FPSIMD
registers since the time the task was last scheduled out.  Such
changes are not guaranteed to be reflected immedately in
thread_struct.

As a result, a coredump can contain stale values for these
registers.  Reading the registers of a stopped task via ptrace is
unaffected.

This patch explicitly flushes the CPU state back to thread_struct
before dumping when operating on current, thus ensuring that
coredump contents are up to date.

Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: ptrace: Fix VFP register dumping in compat coredumps
Dave Martin [Wed, 21 Jun 2017 15:00:42 +0000 (16:00 +0100)]
arm64: ptrace: Fix VFP register dumping in compat coredumps

Currently, VFP registers are omitted from coredumps for compat
processes, due to a bug in the REGSET_COMPAT_VFP regset
implementation.

compat_vfp_get() needs to transfer non-contiguous data from
thread_struct.fpsimd_state, and uses put_user() to handle the
offending trailing word (FPSCR).  This fails when copying to a
kernel address (i.e., kbuf && !ubuf), which is what happens when
dumping core.  As a result, the ELF coredump core code silently
omits the NT_ARM_VFP note from the dump.

It would be possible to work around this with additional special
case code for the put_user(), but since user_regset_copyout() is
explicitly designed to handle this scenario it is cleaner to port
the put_user() to a user_regset_copyout() call, which this patch
does.

Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoefi: parse ARM processor error
Tyler Baicar [Wed, 21 Jun 2017 18:17:07 +0000 (12:17 -0600)]
efi: parse ARM processor error

Add support for ARM Common Platform Error Record (CPER).
UEFI 2.6 specification adds support for ARM specific
processor error information to be reported as part of the
CPER records. This provides more detail on for processor error logs.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Reviewed-by: James Morse <james.morse@arm.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agocper: add timestamp print to CPER status printing
Tyler Baicar [Wed, 21 Jun 2017 18:17:06 +0000 (12:17 -0600)]
cper: add timestamp print to CPER status printing

The ACPI 6.1 spec added a timestamp to the generic error data
entry structure. Print the timestamp out when printing out the
error information.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoras: acpi/apei: cper: add support for generic data v3 structure
Tyler Baicar [Wed, 21 Jun 2017 18:17:05 +0000 (12:17 -0600)]
ras: acpi/apei: cper: add support for generic data v3 structure

The ACPI 6.1 spec adds a new revision of the generic error data
entry structure. Add support to handle the new structure as well
as properly verify and iterate through the generic data entries.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoacpi: apei: read ack upon ghes record consumption
Tyler Baicar [Wed, 21 Jun 2017 18:17:04 +0000 (12:17 -0600)]
acpi: apei: read ack upon ghes record consumption

A RAS (Reliability, Availability, Serviceability) controller
may be a separate processor running in parallel with OS
execution, and may generate error records for consumption by
the OS. If the RAS controller produces multiple error records,
then they may be overwritten before the OS has consumed them.

The Generic Hardware Error Source (GHES) v2 structure
introduces the capability for the OS to acknowledge the
consumption of the error record generated by the RAS
controller. A RAS controller supporting GHESv2 shall wait for
the acknowledgment before writing a new error record, thus
eliminating the race condition.

Add support for parsing of GHESv2 sub-tables as well.

Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
CC: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Reviewed-by: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoMerge branch 'uuid-types' of git://git.infradead.org/users/hch/uuid into aarch64...
Will Deacon [Tue, 20 Jun 2017 18:11:48 +0000 (19:11 +0100)]
Merge branch 'uuid-types' of git://git.infradead.org/users/hch/uuid into aarch64/for-next/ras-apei

Pull in uuid-types branch from Christoph, since this conflicts heavily
with the ACPI/APEI RAS work from Tyler Baicer and was created as an
immutable branch to avoid conflicts with ACPI development.

7 years agoarm64: pass machine size to sparse
Luc Van Oostenryck [Tue, 20 Jun 2017 12:24:43 +0000 (14:24 +0200)]
arm64: pass machine size to sparse

When using sparse on the arm64 tree we get many thousands of
warnings like 'constant ... is so big it is unsigned long long'
or 'shift too big (32) for type unsigned long'. This happens
because by default sparse considers the machine as 32bit and
defines the size of the types accordingly.

Fix this by passing the '-m64' flag to sparse so that
sparse can correctly define longs as being 64bit.

CC: Catalin Marinas <catalin.marinas@arm.com>
CC: Will Deacon <will.deacon@arm.com>
CC: linux-arm-kernel@lists.infradead.org
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: signal: factor out signal frame record allocation
Dave Martin [Thu, 15 Jun 2017 14:03:41 +0000 (15:03 +0100)]
arm64: signal: factor out signal frame record allocation

This patch factors out the allocator for signal frame optional
records into a separate function, to ensure consistency and
facilitate later expansion.

No overrun checking is currently done, because the allocation is in
user memory and anyway the kernel never tries to allocate enough
space in the signal frame yet for an overrun to occur.  This
behaviour will be refined in future patches.

The approach taken in this patch to allocation of the terminator
record is not very clean: this will also be replaced in subsequent
patches.

For future extension, a comment is added in sigcontext.h
documenting the current static allocations in __reserved[].  This
will be important for determining under what circumstances
userspace may or may not see an expanded signal frame.

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: signal: factor frame layout and population into separate passes
Dave Martin [Thu, 15 Jun 2017 14:03:40 +0000 (15:03 +0100)]
arm64: signal: factor frame layout and population into separate passes

In preparation for expanding the signal frame, this patch refactors
the signal frame setup code in setup_sigframe() into two separate
passes.

The first pass, setup_sigframe_layout(), determines the size of the
signal frame and its internal layout, including the presence and
location of optional records.  The resulting knowledge is used to
allocate and locate the user stack space required for the signal
frame and to determine which optional records to include.

The second pass, setup_sigframe(), is called once the stack frame
is allocated in order to populate it with the necessary context
information.

As a result of these changes, it becomes more natural to represent
locations in the signal frame by a base pointer and an offset,
since the absolute address of each location is not known during the
layout pass.  To be more consistent with this logic,
parse_user_sigframe() is refactored to describe signal frame
locations in a similar way.

This change has no effect on the signal ABI, but will make it
easier to expand the signal frame in future patches.

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: signal: Refactor sigcontext parsing in rt_sigreturn
Dave Martin [Thu, 15 Jun 2017 14:03:39 +0000 (15:03 +0100)]
arm64: signal: Refactor sigcontext parsing in rt_sigreturn

Currently, rt_sigreturn does very limited checking on the
sigcontext coming from userspace.

Future additions to the sigcontext data will increase the potential
for surprises.  Also, it is not clear whether the sigcontext
extension records are supposed to occur in a particular order.

To allow the parsing code to be extended more easily, this patch
factors out the sigcontext parsing into a separate function, and
adds extra checks to validate the well-formedness of the sigcontext
structure.

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: signal: split frame link record from sigcontext structure
Dave Martin [Thu, 15 Jun 2017 14:03:38 +0000 (15:03 +0100)]
arm64: signal: split frame link record from sigcontext structure

In order to be able to increase the amount of the data currently
written to the __reserved[] array in the signal frame, it is
necessary to overwrite the locations currently occupied by the
{fp,lr} frame link record pushed at the top of the signal stack.

In order for this to work, this patch detaches the frame link
record from struct rt_sigframe and places it separately at the top
of the signal stack.  This will allow subsequent patches to insert
data between it and __reserved[].

This change relies on the non-ABI status of the placement of the
frame record with respect to struct sigframe: this status is
undocumented, but the placement is not declared or described in the
user headers, and known unwinder implementations (libgcc,
libunwind, gdb) appear not to rely on it.

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: mm: select CONFIG_ARCH_PROC_KCORE_TEXT
Ard Biesheuvel [Wed, 14 Jun 2017 10:43:55 +0000 (12:43 +0200)]
arm64: mm: select CONFIG_ARCH_PROC_KCORE_TEXT

To avoid issues with the /proc/kcore code getting confused about the
kernels block mappings in the VMALLOC region, enable the existing
facility that describes the [_text, _end) interval as a separate
KCORE_TEXT region, which supersedes the KCORE_VMALLOC region that
it intersects with on arm64.

Reported-by: Tan Xiaojun <tanxiaojun@huawei.com>
Tested-by: Tan Xiaojun <tanxiaojun@huawei.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Laura Abbott <labbott@redhat.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agofs/proc: kcore: use kcore_list type to check for vmalloc/module address
Ard Biesheuvel [Wed, 14 Jun 2017 10:43:54 +0000 (12:43 +0200)]
fs/proc: kcore: use kcore_list type to check for vmalloc/module address

Instead of passing each start address into is_vmalloc_or_module_addr()
to decide whether it falls into either the VMALLOC or the MODULES region,
we can simply check the type field of the current kcore_list entry, since
it will be set to KCORE_VMALLOC based on exactly the same conditions.

As a bonus, when reading the KCORE_TEXT region on architectures that have
one, this will avoid using vread() on the region if it happens to intersect
with a KCORE_VMALLOC region. This is due the fact that the KCORE_TEXT
region is the first one to be added to the kcore region list.

Reported-by: Tan Xiaojun <tanxiaojun@huawei.com>
Tested-by: Tan Xiaojun <tanxiaojun@huawei.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Laura Abbott <labbott@redhat.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agodrivers/char: kmem: disable on arm64
Ard Biesheuvel [Tue, 20 Jun 2017 06:59:00 +0000 (08:59 +0200)]
drivers/char: kmem: disable on arm64

As it turns out, arm64 deviates from other architectures in the way it
maps the VMALLOC region: on most (all?) other architectures, it resides
strictly above the kernel's direct mapping of DRAM, but on arm64, this
is the other way around. For instance, for a 48-bit VA configuration,
we have

  modules : 0xffff000000000000 - 0xffff000008000000   (   128 MB)
  vmalloc : 0xffff000008000000 - 0xffff7dffbfff0000   (129022 GB)
  ...
  vmemmap : 0xffff7e0000000000 - 0xffff800000000000   (  2048 GB maximum)
            0xffff7e0000000000 - 0xffff7e0003ff0000   (    63 MB actual)
  memory  : 0xffff800000000000 - 0xffff8000ffc00000   (  4092 MB)

This has mostly gone unnoticed until now, but it does appear that it
breaks an assumption in the kmem read/write code, which does something
like

  if (p < (unsigned long) high_memory) {
    ... use straight copy_[to|from]_user() using p as virtual address ...
  }
  ...
  if (count > 0) {
    ... use vread/vwrite for accesses past high_memory ...
  }

The first condition will inadvertently hold for the VMALLOC region if
VMALLOC_START < PAGE_OFFSET [which is the case on arm64], but the read
or write will subsequently fail the virt_addr_valid() check, resulting
in a -ENXIO return value.

Given how kmem seems to be living in borrowed time anyway, and given
the fact that nobody noticed that the read/write interface is broken
on arm64 in the first place, let's not bother trying to fix it, but
simply disable the /dev/kmem interface entirely for arm64.

Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: Export save_stack_trace_tsk()
Dustin Brown [Tue, 13 Jun 2017 18:40:56 +0000 (11:40 -0700)]
arm64: Export save_stack_trace_tsk()

The kernel watchdog is a great debugging tool for finding tasks that
consume a disproportionate amount of CPU time in contiguous chunks. One
can imagine building a similar watchdog for arbitrary driver threads
using save_stack_trace_tsk() and print_stack_trace(). However, this is
not viable for dynamically loaded driver modules on ARM platforms
because save_stack_trace_tsk() is not exported for those architectures.
Export save_stack_trace_tsk() for the ARM64 architecture to align with
x86 and support various debugging use cases such as arbitrary driver
thread watchdog timers.

Signed-off-by: Dustin Brown <dustinb@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoACPI/IORT: Remove iort_node_match()
Lorenzo Pieralisi [Wed, 14 Jun 2017 16:37:13 +0000 (17:37 +0100)]
ACPI/IORT: Remove iort_node_match()

Commit 316ca8804ea8 ("ACPI/IORT: Remove linker section for IORT entries
probing") removed the linker section for IORT entries probing.

Since those IORT entries were the only iort_node_match() interface
users, the iort_node_match() became obsolete and can then be removed.

Remove the ACPI IORT iort_node_match() interface from the kernel.

Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoARM64/irqchip: Update ACPI_IORT symbol selection logic
Lorenzo Pieralisi [Wed, 14 Jun 2017 16:37:12 +0000 (17:37 +0100)]
ARM64/irqchip: Update ACPI_IORT symbol selection logic

ACPI IORT is an ACPI addendum to describe the connection topology of
devices with IOMMUs and interrupt controllers on ARM64 ACPI systems.

Currently the ACPI IORT Kbuild symbol is selected whenever the Kbuild
symbol ARM_GIC_V3_ITS is enabled, which in turn is selected by ARM64
Kbuild defaults. This makes the logic behind ACPI_IORT selection a bit
twisted and not easy to follow. On ARM64 systems enabling ACPI the
kbuild symbol ACPI_IORT should always be selected in that it is a kernel
layer provided to the ARM64 arch code to parse and enable ACPI firmware
bindings.

Make the ACPI_IORT selection explicit in ARM64 Kbuild and remove the
selection from ARM_GIC_V3_ITS entry, making the ACPI_IORT selection
logic clearer to follow.

Acked-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Hanjun Guo <hanjun.guo@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64/dma-mapping: Remove extraneous null-pointer checks
Olav Haugan [Tue, 13 Jun 2017 20:56:14 +0000 (13:56 -0700)]
arm64/dma-mapping: Remove extraneous null-pointer checks

The current null-pointer check in __dma_alloc_coherent and
__dma_free_coherent is not needed anymore since the
__dma_alloc/__dma_free functions won't be called if !dev (dummy ops will
be called instead).

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Olav Haugan <ohaugan@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agodrivers/perf: commonise PERF_EVENTS dependency
Mark Rutland [Tue, 13 Jun 2017 12:45:51 +0000 (13:45 +0100)]
drivers/perf: commonise PERF_EVENTS dependency

All PMU drivers are going to depend on PERF_EVENTS, so let's make this
dependency common and simplify the individual Kconfig entries.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm: perf: make of_device_ids const
Arvind Yadav [Thu, 15 Jun 2017 09:25:37 +0000 (14:55 +0530)]
arm: perf: make of_device_ids const

of_device_ids are not supposed to change at runtime. All functions
working with of_device_ids provided by <linux/of.h> work with const
of_device_ids. So mark the non-const structs as const.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: kconfig: allow support for memory failure handling
Jonathan (Zhixiong) Zhang [Thu, 8 Jun 2017 17:25:29 +0000 (18:25 +0100)]
arm64: kconfig: allow support for memory failure handling

Declare ARCH_SUPPORTS_MEMORY_FAILURE, as arm64 does support
memory failure recovery attempt.

Signed-off-by: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
(Dropped changes to ACPI APEI Kconfig and updated commit log)
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Steve Capper <steve.capper@arm.com>
Tested-by: Manoj Iyer <manoj.iyer@canonical.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: mm: Update perf accounting to handle poison faults
Punit Agrawal [Thu, 8 Jun 2017 17:25:28 +0000 (18:25 +0100)]
arm64: mm: Update perf accounting to handle poison faults

Re-organise the perf accounting for fault handling in preparation for
enabling handling of hardware poison faults in subsequent commits. The
change updates perf accounting to be inline with the behaviour on
x86.

With this update, the perf fault accounting -

  * Always report PERF_COUNT_SW_PAGE_FAULTS

  * Doesn't report anything else for VM_FAULT_ERROR (which includes
    hwpoison faults)

  * Reports PERF_COUNT_SW_PAGE_FAULTS_MAJ if it's a major
    fault (indicated by VM_FAULT_MAJOR)

  * Otherwise, reports PERF_COUNT_SW_PAGE_FAULTS_MIN

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: hwpoison: add VM_FAULT_HWPOISON[_LARGE] handling
Jonathan (Zhixiong) Zhang [Thu, 8 Jun 2017 17:25:27 +0000 (18:25 +0100)]
arm64: hwpoison: add VM_FAULT_HWPOISON[_LARGE] handling

Add VM_FAULT_HWPOISON[_LARGE] handling to the arm64 page fault
handler. Handling of VM_FAULT_HWPOISON[_LARGE] is very similar
to VM_FAULT_OOM, the only difference is that a different si_code
(BUS_MCEERR_AR) is passed to user space and si_addr_lsb field is
initialized.

Signed-off-by: Jonathan (Zhixiong) Zhang <zjzhang@codeaurora.org>
Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
(fix new __do_user_fault call-site)
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Steve Capper <steve.capper@arm.com>
Tested-by: Manoj Iyer <manoj.iyer@canonical.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: hugetlb: Fix huge_pte_offset to return poisoned page table entries
Punit Agrawal [Thu, 8 Jun 2017 17:25:26 +0000 (18:25 +0100)]
arm64: hugetlb: Fix huge_pte_offset to return poisoned page table entries

When memory failure is enabled, a poisoned hugepage pte is marked as a
swap entry. huge_pte_offset() does not return the poisoned page table
entries when it encounters PUD/PMD hugepages.

This behaviour of huge_pte_offset() leads to error such as below when
munmap is called on poisoned hugepages.

[  344.165544] mm/pgtable-generic.c:33: bad pmd 000000083af00074.

Fix huge_pte_offset() to return the poisoned pte which is then
appropriately handled by the generic layer code.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Steve Capper <steve.capper@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Tested-by: Manoj Iyer <manoj.iyer@canonical.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: ftrace: fix building without CONFIG_MODULES
Will Deacon [Mon, 12 Jun 2017 13:43:25 +0000 (14:43 +0100)]
arm64: ftrace: fix building without CONFIG_MODULES

When CONFIG_MODULES is disabled, we cannot dereference a module pointer:

arch/arm64/kernel/ftrace.c: In function 'ftrace_make_call':
arch/arm64/kernel/ftrace.c:107:36: error: dereferencing pointer to incomplete type 'struct module'
   trampoline = (unsigned long *)mod->arch.ftrace_trampoline;

Also, the within_module() function is not defined:

arch/arm64/kernel/ftrace.c: In function 'ftrace_make_nop':
arch/arm64/kernel/ftrace.c:171:8: error: implicit declaration of function 'within_module'; did you mean 'init_module'? [-Werror=implicit-function-declaration]

This addresses both by adding replacing the IS_ENABLED(CONFIG_ARM64_MODULE_PLTS)
checks with #ifdef versions.

Fixes: e71a4e1bebaf ("arm64: ftrace: add support for far branches to dynamic ftrace")
Reported-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: fault: Print info about page table structure when dumping pte
Will Deacon [Mon, 15 May 2017 14:23:58 +0000 (15:23 +0100)]
arm64: fault: Print info about page table structure when dumping pte

Whilst debugging a remote crash, I noticed that show_pte is unhelpful
when it comes to describing the structure of the page table being walked.
This is easily fixed by printing out the page table (swapper vs user),
page size and virtual address size when displaying the PGD address.

Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: mm: print file name of faulting vma
Kristina Martsenko [Fri, 9 Jun 2017 15:35:54 +0000 (16:35 +0100)]
arm64: mm: print file name of faulting vma

Print out the name of the file associated with the vma that faulted.
This is usually the executable or shared library name. We already print
out the task name, but also printing the library name is useful for
pinpointing bugs to libraries.

Also print the base address and size of the vma, which together with the
PC (printed by __show_regs) gives the offset into the library.

Fault prints now look like:
test[2361]: unhandled level 2 translation fault (11) at 0x00000012, esr 0x92000006, in libfoo.so[ffffa0145000+1000]

This is already done on x86, for more details see commit 03252919b798
("x86: print which shared library/executable faulted in segfault etc.
messages v3").

Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: mm: don't print out page table entries on EL0 faults
Kristina Martsenko [Fri, 9 Jun 2017 15:35:53 +0000 (16:35 +0100)]
arm64: mm: don't print out page table entries on EL0 faults

When we take a fault from EL0 that can't be handled, we print out the
page table entries associated with the faulting address. This allows
userspace to print out any current page table entries, including kernel
(TTBR1) entries. Exposing kernel mappings like this could pose a
security risk, so don't print out page table information on EL0 faults.
(But still print it out for EL1 faults.) This also follows the same
behaviour as x86, printing out page table entries on kernel mode faults
but not user mode faults.

Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: mm: print out correct page table entries
Kristina Martsenko [Fri, 9 Jun 2017 15:35:52 +0000 (16:35 +0100)]
arm64: mm: print out correct page table entries

When we take a fault that can't be handled, we print out the page table
entries associated with the faulting address. In some cases we currently
print out the wrong entries. For a faulting TTBR1 address, we sometimes
print out TTBR0 table entries instead, and for a faulting TTBR0 address
we sometimes print out TTBR1 table entries. Fix this by choosing the
tables based on the faulting address.

Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com>
[will: zero-extend addrs to 64-bit, don't walk swapper w/ TTBR0 addr]
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agothermal: int340x_thermal: fix compile after the UUID API switch
Christoph Hellwig [Fri, 9 Jun 2017 14:36:50 +0000 (16:36 +0200)]
thermal: int340x_thermal: fix compile after the UUID API switch

Fix the compile after the switch to the UUID API in commit f4c19ac9
("thermal: int340x_thermal: Switch to use new generic UUID API").

Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agothermal: int340x_thermal: Switch to use new generic UUID API
Andy Shevchenko [Fri, 9 Jun 2017 08:33:06 +0000 (11:33 +0300)]
thermal: int340x_thermal: Switch to use new generic UUID API

There are new types and helpers that are supposed to be used in
new code.

As a preparation to get rid of legacy types and API functions do
the conversion here.

The conversion fixes a potential bug in int340x_thermal as well
since we have to use memcmp() on binary data.

Acked-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoacpi: always include uuid.h
Christoph Hellwig [Thu, 8 Jun 2017 07:02:20 +0000 (09:02 +0200)]
acpi: always include uuid.h

Without this the build will fail for !CONFIG_ACPI builds on x86.

Fixes: 94116f81 ("ACPI: Switch to use generic guid_t in acpi_evaluate_dsm()")
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoarm64: ftrace: add support for far branches to dynamic ftrace
Ard Biesheuvel [Tue, 6 Jun 2017 17:00:22 +0000 (17:00 +0000)]
arm64: ftrace: add support for far branches to dynamic ftrace

Currently, dynamic ftrace support in the arm64 kernel assumes that all
core kernel code is within range of ordinary branch instructions that
occur in module code, which is usually the case, but is no longer
guaranteed now that we have support for module PLTs and address space
randomization.

Since on arm64, all patching of branch instructions involves function
calls to the same entry point [ftrace_caller()], we can emit the modules
with a trampoline that has unlimited range, and patch both the trampoline
itself and the branch instruction to redirect the call via the trampoline.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
[will: minor clarification to smp_wmb() comment]
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoarm64: ftrace: don't validate branch via PLT in ftrace_make_nop()
Ard Biesheuvel [Tue, 6 Jun 2017 17:00:21 +0000 (17:00 +0000)]
arm64: ftrace: don't validate branch via PLT in ftrace_make_nop()

When turning branch instructions into NOPs, we attempt to validate the
action by comparing the old value at the call site with the opcode of
a direct relative branch instruction pointing at the old target.

However, these call sites are statically initialized to call _mcount(),
and may be redirected via a PLT entry if the module is loaded far away
from the kernel text, leading to false negatives and spurious errors.

So skip the validation if CONFIG_ARM64_MODULE_PLTS is configured.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoACPI: Switch to use generic guid_t in acpi_evaluate_dsm()
Andy Shevchenko [Mon, 5 Jun 2017 16:40:46 +0000 (19:40 +0300)]
ACPI: Switch to use generic guid_t in acpi_evaluate_dsm()

acpi_evaluate_dsm() and friends take a pointer to a raw buffer of 16
bytes. Instead we convert them to use guid_t type. At the same time we
convert current users.

acpi_str_to_uuid() becomes useless after the conversion and it's safe to
get rid of it.

Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Cc: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Acked-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Acked-by: Joerg Roedel <jroedel@suse.de>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Yisen Zhuang <yisen.zhuang@huawei.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Acked-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Acked-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoarm64, vdso: Define vdso_{start,end} as array
Kees Cook [Tue, 6 Jun 2017 04:52:30 +0000 (21:52 -0700)]
arm64, vdso: Define vdso_{start,end} as array

Adjust vdso_{start|end} to be char arrays to avoid compile-time analysis
that flags "too large" memcmp() calls with CONFIG_FORTIFY_SOURCE.

Cc: Jisheng Zhang <jszhang@marvell.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoACPI / extlog: Switch to use new generic UUID API
Andy Shevchenko [Mon, 5 Jun 2017 16:40:45 +0000 (19:40 +0300)]
ACPI / extlog: Switch to use new generic UUID API

There are new types and helpers that are supposed to be used in new code.

As a preparation to get rid of legacy types and API functions do
the conversion here.

Cc: Borislav Petkov <bp@suse.de>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoACPI / bus: Switch to use new generic UUID API
Andy Shevchenko [Mon, 5 Jun 2017 16:40:44 +0000 (19:40 +0300)]
ACPI / bus: Switch to use new generic UUID API

There are new types and helpers that are supposed to be used in new code.

As a preparation to get rid of legacy types and API functions do
the conversion here.

Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoACPI / APEI: Switch to use new generic UUID API
Andy Shevchenko [Mon, 5 Jun 2017 16:40:43 +0000 (19:40 +0300)]
ACPI / APEI: Switch to use new generic UUID API

There are new types and helpers that are supposed to be used in new code.

As a preparation to get rid of legacy types and API functions do
the conversion here.

Cc: Borislav Petkov <bp@suse.de>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoacpi, nfit: Switch to use new generic UUID API
Andy Shevchenko [Mon, 5 Jun 2017 16:40:42 +0000 (19:40 +0300)]
acpi, nfit: Switch to use new generic UUID API

There are new types and helpers that are supposed to be used in new code.

As a preparation to get rid of legacy types and API functions do
the conversion here.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoMAINTAINERS: add uuid entry
Christoph Hellwig [Tue, 23 May 2017 08:52:58 +0000 (10:52 +0200)]
MAINTAINERS: add uuid entry

I'll keep maintaining whatever little changed we need here, with Andy as
my designated reviewer.

Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agotmpfs: generate random sb->s_uuid
Amir Goldstein [Thu, 18 May 2017 12:29:33 +0000 (15:29 +0300)]
tmpfs: generate random sb->s_uuid

This is used by overlayfs to encode intrasystem unique file handles.

Suggested-by: Miklos Szeredi <mszeredi@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoscsi_debug: switch to uuid_t
Christoph Hellwig [Wed, 17 May 2017 07:55:26 +0000 (09:55 +0200)]
scsi_debug: switch to uuid_t

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agonvme: switch to uuid_t
Christoph Hellwig [Wed, 17 May 2017 07:54:27 +0000 (09:54 +0200)]
nvme: switch to uuid_t

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agosysctl: switch to use uuid_t
Christoph Hellwig [Wed, 17 May 2017 07:51:09 +0000 (09:51 +0200)]
sysctl: switch to use uuid_t

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agopartitions/ldm: switch to use uuid_t
Christoph Hellwig [Wed, 17 May 2017 07:38:37 +0000 (09:38 +0200)]
partitions/ldm: switch to use uuid_t

And the uuid helpers.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agooverlayfs: use uuid_t instead of uuid_be
Christoph Hellwig [Wed, 17 May 2017 07:32:50 +0000 (09:32 +0200)]
overlayfs: use uuid_t instead of uuid_be

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agofs: switch ->s_uuid to uuid_t
Christoph Hellwig [Wed, 10 May 2017 13:06:33 +0000 (15:06 +0200)]
fs: switch ->s_uuid to uuid_t

For some file systems we still memcpy into it, but in various places this
already allows us to use the proper uuid helpers.  More to come..

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com> (Changes to IMA/EVM)
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agoima/policy: switch to use uuid_t
Christoph Hellwig [Thu, 1 Jun 2017 05:00:26 +0000 (07:00 +0200)]
ima/policy: switch to use uuid_t

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agoblock: remove blk_part_pack_uuid
Christoph Hellwig [Wed, 10 May 2017 13:16:44 +0000 (15:16 +0200)]
block: remove blk_part_pack_uuid

This helper was only used by IMA of all things, which would get spurious
errors if CONFIG_BLOCK is disabled.  Just opencode the call there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agoxfs: use the common helper uuid_is_null()
Amir Goldstein [Thu, 4 May 2017 13:26:23 +0000 (16:26 +0300)]
xfs: use the common helper uuid_is_null()

Use the common helper uuid_is_null() and remove the xfs specific
helper uuid_is_nil().

The common helper does not check for the NULL pointer value as
xfs helper did, but xfs code never calls the helper with a pointer
that can be NULL.

Conform comments and warning strings to use the term 'null uuid'
instead of 'nil uuid', because this is the terminology used by
lib/uuid.c and its users. It is also the terminology used in
userspace by libuuid and xfsprogs.

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
[hch: remove now unused uuid.[ch]]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agoxfs: remove uuid_getnodeuniq and xfs_uu_t
Christoph Hellwig [Fri, 5 May 2017 07:39:10 +0000 (09:39 +0200)]
xfs: remove uuid_getnodeuniq and xfs_uu_t

Opencode uuid_getnodeuniq in the only caller, and directly decode
the uuid_t representation instead of using a structure cast for it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agoS390/sysinfo: use uuid_is_null instead of opencoding it
Christoph Hellwig [Thu, 11 May 2017 12:00:57 +0000 (14:00 +0200)]
S390/sysinfo: use uuid_is_null instead of opencoding it

And switch to use uuid_t instead of the old uuid_be type.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agouuid: hoist uuid_is_null() helper from libnvdimm
Christoph Hellwig [Thu, 11 May 2017 07:01:42 +0000 (09:01 +0200)]
uuid: hoist uuid_is_null() helper from libnvdimm

Hoist the libnvdimm helper as an inline helper to linux/uuid.h
using an auxiliary const variable uuid_null in lib/uuid.c.

[hch: also add the guid variant.  Both do the same but I'd like
to keep casts to a minimum]

The common helper uses the new abstract type uuid_t * instead of
u8 *.

Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
[hch: added guid_is_null]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agouuid: hoist helpers uuid_equal() and uuid_copy() from xfs
Christoph Hellwig [Thu, 11 May 2017 12:00:57 +0000 (14:00 +0200)]
uuid: hoist helpers uuid_equal() and uuid_copy() from xfs

These helper are used to compare and copy two uuid_t type objects.

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
[hch: also provide the respective guid_ versions]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agouuid: don't export guid_index and uuid_index
Christoph Hellwig [Wed, 10 May 2017 12:22:11 +0000 (14:22 +0200)]
uuid: don't export guid_index and uuid_index

These are only used in uuid.c and vsprintf.c and aren't something modules
should use directly.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agouuid: rename uuid types
Christoph Hellwig [Wed, 17 May 2017 08:02:48 +0000 (10:02 +0200)]
uuid: rename uuid types

Our "little endian" UUID really is a Wintel GUID, so rename it and its
helpers such (guid_t).  The big endian UUID is the only true one, so
give it the name uuid_t.  The uuid_le and uuid_be names are retained for
now, but will hopefully go away soon.  The exception to that are the _cmp
helpers that will be replaced by better primitives ASAP and thus don't
get the new names.

Also the _to_bin helpers are named to match the better named uuid_parse
routine in userspace.

Also remove the existing typedef in XFS that's now been superceeded by
the generic type name.

Signed-off-by: Christoph Hellwig <hch@lst.de>
[andy: also update the UUID_LE/UUID_BE macros including fallout]
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agouuid: remove uuid_be defintions from the uapi header
Christoph Hellwig [Wed, 17 May 2017 07:56:45 +0000 (09:56 +0200)]
uuid: remove uuid_be defintions from the uapi header

We don't use uuid_be and the UUID_BE constants in any uapi headers, so make
them private to the kernel.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agonfsd: namespace-prefix uuid_parse
Christoph Hellwig [Wed, 31 May 2017 14:40:52 +0000 (16:40 +0200)]
nfsd: namespace-prefix uuid_parse

Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agomd: namespace private helper names
Amir Goldstein [Thu, 4 May 2017 13:26:20 +0000 (16:26 +0300)]
md: namespace private helper names

The md private helper uuid_equal() collides with a generic helper
of the same name.

Rename the md private helper to md_uuid_equal() and do the same for
md_sb_equal().

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Shaohua Li <shli@kernel.org>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
7 years agoxfs: use uuid_be to implement the uuid_t type
Christoph Hellwig [Fri, 5 May 2017 07:53:09 +0000 (09:53 +0200)]
xfs: use uuid_be to implement the uuid_t type

Use the generic Linux definition to implement our UUID type, this will
allow using more generic infrastructure in the future.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
7 years agoxfs: use uuid_copy() helper to abstract uuid_t
Amir Goldstein [Thu, 4 May 2017 13:26:16 +0000 (16:26 +0300)]
xfs: use uuid_copy() helper to abstract uuid_t

uuid_t definition is about to change.

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
7 years agouuid,afs: move struct uuid_v1 back into afs
Christoph Hellwig [Sun, 28 May 2017 05:56:46 +0000 (08:56 +0300)]
uuid,afs: move struct uuid_v1 back into afs

This essentially is a partial revert of commit ff548773
("afs: Move UUID struct to linux/uuid.h") and moves struct uuid_v1 back into
fs/afs as struct afs_uuid.  It however keeps it as big endian structure
so that we can use the normal uuid generation helpers when casting to/from
struct afs_uuid.

The V1 uuid intrepretation in struct form isn't really useful to the
rest of the kernel, and not really compatible to it either, so move it
back to AFS instead of polluting the global uuid.h.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: David Howells <dhowells@redhat.com>
7 years agoarm64: cpufeature: Fix CPU_OUT_OF_SPEC taint for uniform systems
Will Deacon [Mon, 5 Jun 2017 10:40:23 +0000 (11:40 +0100)]
arm64: cpufeature: Fix CPU_OUT_OF_SPEC taint for uniform systems

Commit 3fde2999fac5 ("arm64: cpufeature: Don't dump useless backtrace on
CPU_OUT_OF_SPEC") changed the cpufeature detection code to use add_taint
instead of WARN_TAINT_ONCE when detecting a heterogeneous system with
mismatched feature support. Unfortunately, this resulted in all systems
getting the taint, regardless of any feature mismatch.

This patch fixes the problem by conditionalising the taint on detecting
a feature mismatch.

Acked-by: Mark Rutland <mark.rutland@arm.com>
Reported-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
7 years agoLinux 4.12-rc4
Linus Torvalds [Sun, 4 Jun 2017 23:47:43 +0000 (16:47 -0700)]
Linux 4.12-rc4

7 years agofs/ufs: Set UFS default maximum bytes per file
Richard Narron [Sun, 4 Jun 2017 23:23:18 +0000 (16:23 -0700)]
fs/ufs: Set UFS default maximum bytes per file

This fixes a problem with reading files larger than 2GB from a UFS-2
file system:

    https://bugzilla.kernel.org/show_bug.cgi?id=195721

The incorrect UFS s_maxsize limit became a problem as of commit
c2a9737f45e2 ("vfs,mm: fix a dead loop in truncate_inode_pages_range()")
which started using s_maxbytes to avoid a page index overflow in
do_generic_file_read().

That caused files to be truncated on UFS-2 file systems because the
default maximum file size is 2GB (MAX_NON_LFS) and UFS didn't update it.

Here I simply increase the default to a common value used by other file
systems.

Signed-off-by: Richard Narron <comet.berkeley@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Will B <will.brokenbourgh2877@gmail.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: <stable@vger.kernel.org> # v4.9 and backports of c2a9737f45e2
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge tag 'nfs-for-4.12-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Sun, 4 Jun 2017 18:56:53 +0000 (11:56 -0700)]
Merge tag 'nfs-for-4.12-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:
 "Bugfixes include:

   - Fix a typo in commit e092693443b ("NFS append COMMIT after
     synchronous COPY") that breaks copy offload

   - Fix the connect error propagation in xs_tcp_setup_socket()

   - Fix a lock leak in nfs40_walk_client_list

   - Verify that pNFS requests lie within the offset range of the layout
     segment"

* tag 'nfs-for-4.12-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  nfs: Mark unnecessarily extern functions as static
  SUNRPC: ensure correct error is reported by xs_tcp_setup_socket()
  NFSv4.0: Fix a lock leak in nfs40_walk_client_list
  pnfs: Fix the check for requests in range of layout segment
  xprtrdma: Delete an error message for a failed memory allocation in xprt_rdma_bc_setup()
  pNFS/flexfiles: missing error code in ff_layout_alloc_lseg()
  NFS fix COMMIT after COPY

7 years agoMerge tag 'tty-4.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sun, 4 Jun 2017 18:41:41 +0000 (11:41 -0700)]
Merge tag 'tty-4.12-rc4' of git://git./linux/kernel/git/gregkh/tty

Pull tty fix from Greg KH:
 "Here is a single tty core fix for 4.12-rc4. It reverts a patch that a
  lot of people reported as causing lockdep and other warnings.

  Right after I reverted this in my tree, it seems like another
  "correct" fix might have shown up, but it's too late in the release
  cycle to be messing with tty core locking, so let's just revert this
  for now to go back how things always have been and try it again for
  4.13.

  This has not been in linux-next as I only reverted it a few hours ago"

* tag 'tty-4.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  Revert "tty: fix port buffer locking"

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sun, 4 Jun 2017 18:37:42 +0000 (11:37 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input subsystem fixes from Dmitry Torokhov:

 - a couple of regression fixes in synaptics and axp20x-pek drivers

 - try to ease transition from PS/2 to RMI for Synaptics touchpad users
   by ensuring we do not try to activate RMI mode when RMI SMBus support
   is not enabled, and nag users a bit to enable it

 - plus a couple of other changes that seemed worthwhile for this
   release

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: axp20x-pek - switch to acpi_dev_present and check for ACPI0011 too
  Input: axp20x-pek - only check for "INTCFD9" ACPI device on Cherry Trail
  Input: tm2-touchkey - use LEN_ON as boolean value instead of LED_FULL
  Input: synaptics - tell users to report when they should be using rmi-smbus
  Input: synaptics - warn the users when there is a better mode
  Input: synaptics - keep PS/2 around when RMI4_SMB is not enabled
  Input: synaptics - clear device info before filling in
  Input: silead - disable interrupt during suspend

7 years agoMerge tag 'rtc-4.12-2' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni...
Linus Torvalds [Sun, 4 Jun 2017 18:29:32 +0000 (11:29 -0700)]
Merge tag 'rtc-4.12-2' of git://git./linux/kernel/git/abelloni/linux

Pull RTC fixlet from Alexandre Belloni:
 "A single patch, not really a fix but I don't think there is any reason
  to delay it.

  Change the mailing list address"

* tag 'rtc-4.12-2' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
  MAINTAINERS: update RTC mailing list

7 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sun, 4 Jun 2017 18:15:43 +0000 (11:15 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "This is nine fixes, seven of which are for the qedi driver (new as of
  4.10) the other two are a use after free in the cxgbi drivers and a
  potential NULL dereference in the rdac device handler"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: libcxgbi: fix skb use after free
  scsi: qedi: Fix endpoint NULL panic during recovery.
  scsi: qedi: set max_fin_rt default value
  scsi: qedi: Set firmware tcp msl timer value.
  scsi: qedi: Fix endpoint NULL panic in qedi_set_path.
  scsi: qedi: Set dma_boundary to 0xfff.
  scsi: qedi: Correctly set firmware max supported BDs.
  scsi: qedi: Fix bad pte call trace when iscsiuio is stopped.
  scsi: scsi_dh_rdac: Use ctlr directly in rdac_failover_get()