openwrt/staging/blogic.git
6 years agovmw_balloon: stats rework
Nadav Amit [Thu, 20 Sep 2018 17:30:17 +0000 (10:30 -0700)]
vmw_balloon: stats rework

To allow the balloon statistics to be updated concurrently, we change
the statistics to be held per core and aggregate it when needed.

To avoid the memory overhead of keeping the statistics per core, and
since it is likely not used by most users, we start updating the
statistics only after the first use. A read-write semaphore is used to
protect the statistics initialization and avoid races. This semaphore is
(and will) be used to protect configuration changes during reset.

While we are at it, address some other issues: change the statistics
update to inline functions instead of define; use ulong for saving the
statistics; and clean the statistics printouts.

Note that this patch changes the format of the outputs. If there are any
automatic tools that use the statistics, they might fail.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: simplify vmballoon_send_get_target()
Nadav Amit [Thu, 20 Sep 2018 17:30:16 +0000 (10:30 -0700)]
vmw_balloon: simplify vmballoon_send_get_target()

As we want to leave as little as possible on the global balloon
structure, to avoid possible future races, we want to get rid sysinfo.
We can actually get the total_ram directly, and simplify the logic of
vmballoon_send_get_target() a little.

While we are doing that, let's return int and avoid mistakes due to
bool/int conversions.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: refactor change size from vmballoon_work
Nadav Amit [Thu, 20 Sep 2018 17:30:15 +0000 (10:30 -0700)]
vmw_balloon: refactor change size from vmballoon_work

The required change in the balloon size is currently computed in
vmballoon_work(), vmballoon_inflate() and vmballoon_deflate(). Refactor
it to simplify the next patches.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: rename VMW_BALLOON_2M_SHIFT to VMW_BALLOON_2M_ORDER
Nadav Amit [Thu, 20 Sep 2018 17:30:14 +0000 (10:30 -0700)]
vmw_balloon: rename VMW_BALLOON_2M_SHIFT to VMW_BALLOON_2M_ORDER

The name of the macro'd VMW_BALLOON_2M_SHIFT is misleading. The value
reflects 2M huge-page order. Unfortunately, we cannot use
HPAGE_PMD_ORDER, since it is not defined when transparent huge-pages are
off, so we need to define our own one.

Rename it to VMW_BALLOON_2M_ORDER. No functional change.

Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: treat all refused pages equally
Nadav Amit [Thu, 20 Sep 2018 17:30:13 +0000 (10:30 -0700)]
vmw_balloon: treat all refused pages equally

Currently, when the hypervisor rejects a page during lock operation, the
VM treats pages differently according to the error-code: in certain
cases the page is immediately freed, and in others it is put on a
rejection list and only freed later.

The behavior does not make too much sense. If the page is freed
immediately it is very likely to be used again in the next batch of
allocations, and be rejected again.

In addition, for support of compaction and OOM notifiers, we wish to
separate the logic that communicates with the hypervisor (as well as
analyzes the status of each page) from the logic that allocates or free
pages.

Treat all errors the same way, queuing the pages on the refuse list.
Move to the next allocation size (4k) when too many pages are refused.
Free the refused pages when moving to the next size to avoid situations
in which too much memory is waiting to be freed on the refused list.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: change batch/single lock abstractions
Nadav Amit [Thu, 20 Sep 2018 17:30:12 +0000 (10:30 -0700)]
vmw_balloon: change batch/single lock abstractions

The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).

The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.

Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.

To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.

Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: remove sleeping allocations
Nadav Amit [Thu, 20 Sep 2018 17:30:11 +0000 (10:30 -0700)]
vmw_balloon: remove sleeping allocations

Splitting the allocations between sleeping and non-sleeping made some
sort of sense as long as rate-limiting was enabled. Now that it is
removed, we need to decide - either we want sleeping allocations or not.

Since no other Linux balloon driver (hv, Xen, virtio) uses sleeping
allocations, use the same approach.

We do distinguish, however, between 2MB allocations and 4kB allocations
and prevent reclamation on 2MB. In both cases, we avoid using emergency
low-memory pools, as it may cause undesired effects.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: simplifying batch access
Nadav Amit [Thu, 20 Sep 2018 17:30:10 +0000 (10:30 -0700)]
vmw_balloon: simplifying batch access

The use of accessors for batch entries complicates the code and makes it
less readable. Remove it an instead use bit-fields.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: merge send_lock and send_unlock path
Nadav Amit [Thu, 20 Sep 2018 17:30:09 +0000 (10:30 -0700)]
vmw_balloon: merge send_lock and send_unlock path

The lock and unlock code paths are very similar, so avoid the duplicate
code by merging them together.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: unify commands tracing and stats
Nadav Amit [Thu, 20 Sep 2018 17:30:08 +0000 (10:30 -0700)]
vmw_balloon: unify commands tracing and stats

Now that we have a single point, unify the tracing and collecting the
statistics for commands and their failure. While it might somewhat
reduce the control over debugging, it cleans the code a lot.

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agovmw_balloon: handle commands in a single function.
Nadav Amit [Thu, 20 Sep 2018 17:30:07 +0000 (10:30 -0700)]
vmw_balloon: handle commands in a single function.

By inlining the hypercall interface, we can unify several operations
into one central point in the code:

- Updating the target.
- Updating when a reset is needed.
- Update statistics (which will be done later in the patch-set).
- Print debug-messages (although they cannot be enabled as selectively).

Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Remove redundant null pointer check before of_node_put and put_device
zhong jiang [Thu, 20 Sep 2018 19:18:19 +0000 (13:18 -0600)]
coresight: Remove redundant null pointer check before of_node_put and put_device

of_node_put and put_device has taken the null pointer check into account.
So it is safe to remove the duplicated check.

Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: dts: binding: Update coresight binding examples
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:18 +0000 (13:18 -0600)]
coresight: dts: binding: Update coresight binding examples

While we updated the coresight DT bindings, some of the
new examples were not updated due to the order in which they
were merged. Let us update all the missed out ones to the
new bindings to avoid confusion.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Rob Herring <robh@kernel.org>
Cc: Frank Rowand <frowand.list@gmail.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: dts: binding: Fix example for TPIU component
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:17 +0000 (13:18 -0600)]
coresight: dts: binding: Fix example for TPIU component

TPIU component has an input port. The example uses out-ports
which is wrong. Let us fix it.

Reported-by: Leo Yan <leo.yan@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: dynamic-replicator: Claim device for use
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:15 +0000 (13:18 -0600)]
coresight: dynamic-replicator: Claim device for use

Use CLAIM protocol to make sure the device is available for use.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: catu: Claim device before use
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:14 +0000 (13:18 -0600)]
coresight: catu: Claim device before use

Use the CLAIM protocol to grab the ownership of the component when
in use.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: funnel: Claim devices before use
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:13 +0000 (13:18 -0600)]
coresight: funnel: Claim devices before use

Use the CLAIM protocol to grab the ownership of the component.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etmx: Claim devices before use
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:12 +0000 (13:18 -0600)]
coresight: etmx: Claim devices before use

Use the CLAIM tags to grab the device for self-hosted usage.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Add support for CLAIM tag protocol
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:11 +0000 (13:18 -0600)]
coresight: Add support for CLAIM tag protocol

Coresight architecture defines CLAIM tags for a device to negotiate
control of the components (external agent vs self-hosted). Each device
has a pair of registers (CLAIMSET & CLAIMCLR) for managing the CLAIM
tags. However, the protocol for the CLAIM tags is IMPLEMENTATION DEFINED.
PSCI has recommendations for the use of the CLAIM tags to negotiate
controls for external agent vs self-hosted use. This patch implements
the recommended protocol by PSCI.

The claim/disclaim operations are performed from the device specific
drivers. The disadvantage is that the calls are sprinkled in each driver,
but this makes the operation much simpler.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: dynamic-replicator: Handle multiple connections
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:10 +0000 (13:18 -0600)]
coresight: dynamic-replicator: Handle multiple connections

When a replicator port is enabled, we block the traffic
on the other port and route all traffic to the new enabled
port. If there are two active trace sessions each targeting
the two different paths from the replicator, the second session
will disable the first session and route all the data to the
second path.
                    ETR
                 /
e.g, replicator
                 \
                    ETB

If CPU0 is operated in sysfs mode to ETR and CPU1 is operated
in perf mode to ETB, depending on the order in which the
replicator is enabled one device is blocked.

Ideally we need trace-id for the session to make the
right choice. That implies we need a trace-id allocation
logic for the coresight subsystem and use that to route
the traffic. The short term solution is to only manage
the "target port" and leave the other port untouched.
That leaves both the paths unaffected, except that some
unwanted traffic may be pushed to the paths (if the Trace-IDs
are not far enough), which is still fine and can be filtered
out while processing rather than silently blocking the data.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etb10: Handle errors enabling the device
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:09 +0000 (13:18 -0600)]
coresight: etb10: Handle errors enabling the device

Prepare the etb10 driver to return errors in enabling
the device.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etm3: Add support for handling errors
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:08 +0000 (13:18 -0600)]
coresight: etm3: Add support for handling errors

Add support for reporting errors back from the SMP cross
function call for enabling ETM.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etm4x: Add support for handling errors
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:07 +0000 (13:18 -0600)]
coresight: etm4x: Add support for handling errors

Add support for handling errors in enabling the component.
The ETM is enabled via cross call to owner CPU. Make
necessary changes to report the error back from the cross
call.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc-etb/etf: Prepare to handle errors enabling
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:06 +0000 (13:18 -0600)]
coresight: tmc-etb/etf: Prepare to handle errors enabling

Prepare to handle errors in enabling the hardware and
report it back to the core driver.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc-etr: Handle errors enabling CATU
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:05 +0000 (13:18 -0600)]
coresight: tmc-etr: Handle errors enabling CATU

Make sure we honor the errors in CATU device and abort the operation.
While at it, delay setting the etr_buf for the session until we are
sure that we are indeed enabling the ETR.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc-etr: Refactor for handling errors
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:04 +0000 (13:18 -0600)]
coresight: tmc-etr: Refactor for handling errors

Refactor the tmc-etr enable operation to make it easier to
handle errors in enabling the hardware. We need to make
sure that the buffer is compatible with the ETR. This
patch re-arranges to make the error handling easier, by
deferring the hardware enablement until all the errors
are checked. This also avoids turning the CATU on/off
during a sysfs read session.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Handle failures in enabling a trace path
Suzuki K Poulose [Thu, 20 Sep 2018 19:18:03 +0000 (13:18 -0600)]
coresight: Handle failures in enabling a trace path

coresight_enable_path() enables the components in a trace
path from a given source to a sink, excluding the source.
The operation is performed in the reverse order; the sink
first and then backwards in the list. However, if we encounter
an error in enabling any of the component, we simply disable
all the components in the given path irrespective of whether
we enabled some of the components in the enable iteration.
This could interfere with another trace session if one of the
link devices is turned off (e.g, TMC-ETF). So, we need to
make sure that we only disable those components which were
actually enabled from the iteration.

This patch achieves the same by refactoring the coresight_disable_path
to accept a "node" to start from in the forward order, which can
then be used from the error path of coresight_enable_path().
With this change, we don't issue a disable call back for a component
which didn't get enabled. This change of behavior triggers
a bug in coresight_enable_link(), where we leave the refcount
on the device and will prevent the device from being enabled
forever. So, we also drop the refcount in the coresight_enable_link()
if the operation failed.

Also, with the refactoring, we always start after the first node (which
is the "SOURCE" device) for disabling the entire path. This implies,
we must not find a "SOURCE" in the middle of the path. Hence, added
a WARN_ON() to make sure the paths we get are sane, rather than
simply ignoring them.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc: Fix byte-address alignment for RRP
Leo Yan [Thu, 20 Sep 2018 19:18:02 +0000 (13:18 -0600)]
coresight: tmc: Fix byte-address alignment for RRP

>From the comment in the code, it claims the requirement for byte-address
alignment for RRP register: 'for 32-bit, 64-bit and 128-bit wide trace
memory, the four LSBs must be 0s. For 256-bit wide trace memory, the
five LSBs must be 0s'.  This isn't consistent with the program, the
program sets five LSBs as zeros for 32/64/128-bit wide trace memory and
set six LSBs zeros for 256-bit wide trace memory.

After checking with the CoreSight Trace Memory Controller technical
reference manual (ARM DDI 0461B, section 3.3.4 RAM Read Pointer
Register), it proves the comment is right and the program does wrong
setting.

This patch fixes byte-address alignment for RRP by following correct
definition in the technical reference manual.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc: Refactor loops in etb dump
Leo Yan [Thu, 20 Sep 2018 19:18:01 +0000 (13:18 -0600)]
coresight: tmc: Refactor loops in etb dump

In ETB dump function tmc_etb_dump_hw() it has nested loops.  The second
level loop is to iterate index in the range [0 .. drvdata->memwidth);
but the index isn't really used in the code, thus the second level
loop is useless.

This patch is to remove the second level loop; the refactor also reduces
indentation and we can use 'break' to replace 'goto' tag.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etm4x: Configure EL2 exception level when kernel is running in HYP
Tomasz Nowicki [Thu, 20 Sep 2018 19:18:00 +0000 (13:18 -0600)]
coresight: etm4x: Configure EL2 exception level when kernel is running in HYP

For non-VHE systems host kernel runs at EL1 and jumps to EL2 whenever
hypervisor code should be executed. In this case ETM4x driver must
restrict configuration to EL1 when it setups kernel tracing.
However, there is no separate hypervisor privilege level when VHE
is enabled, the host kernel runs at EL2.

This patch fixes configuration of TRCACATRn register for VHE systems
so that ETM_EXLEVEL_NS_HYP bit is used instead of ETM_EXLEVEL_NS_OS
to on/off kernel tracing. At the same time, it moves common code
to new helper.

Signed-off-by: Tomasz Nowicki <tnowicki@caviumnetworks.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etb10: Splitting function etb_enable()
Mathieu Poirier [Thu, 20 Sep 2018 19:17:59 +0000 (13:17 -0600)]
coresight: etb10: Splitting function etb_enable()

Up until now the relative simplicity of enabling the ETB made it
possible to accommodate processing for both sysFS and perf methods.
But work on claimtags and CPU-wide trace scenarios is adding some
complexity, making the current code messy and hard to maintain.

As such follow what has been done for ETF and ETR components and split
function etb_enable() so that processing for both API can be done
cleanly.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etb10: Refactor etb_drvdata::mode handling
Mathieu Poirier [Thu, 20 Sep 2018 19:17:58 +0000 (13:17 -0600)]
coresight: etb10: Refactor etb_drvdata::mode handling

This patch moves the etb_drvdata::mode from a locat_t to a simple u32,
as it is for the ETF and ETR drivers.  This streamlines the code and adds
commonality with the other drivers when dealing with similar operations.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etm-perf: Add support for ETR backend
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:57 +0000 (13:17 -0600)]
coresight: etm-perf: Add support for ETR backend

Add support for using TMC-ETR as backend for ETM perf tracing.
We use software double buffering at the moment. i.e, the TMC-ETR
uses a separate buffer than the perf ring buffer. The data is
copied to the perf ring buffer once a session completes.

The TMC-ETR would try to match the larger of perf ring buffer
or the ETR buffer size configured via sysfs, scaling down to
a minimum limit of 1MB.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Remove set_buffer call back
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:56 +0000 (13:17 -0600)]
coresight: perf: Remove set_buffer call back

In coresight perf mode, we need to prepare the sink before
starting a session, which is done via set_buffer call back.
We then proceed to enable the tracing. If we fail to start
the session successfully, we leave the sink configuration
unchanged.  In order to make the operation atomic and to
avoid yet another call back to clear the buffer, we get
rid of the "set_buffer" call back and pass the buffer details
via enable() call back to the sink.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Add helper to retrieve sink configuration
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:55 +0000 (13:17 -0600)]
coresight: perf: Add helper to retrieve sink configuration

We can always find the sink configuration for a given perf_output_handle.
Add a helper to retrieve the sink configuration for a given
perf_output_handle. This will be used to get rid of the set_buffer()
call back.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Remove reset_buffer call back for sinks
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:54 +0000 (13:17 -0600)]
coresight: perf: Remove reset_buffer call back for sinks

Right now we issue an update_buffer() and reset_buffer() call backs
in succession when we stop tracing an event. The update_buffer is
supposed to check the status of the buffer and make sure the ring buffer
is updated with the trace data. And we store information about the
size of the data collected only to be consumed by the reset_buffer
callback which always follows the update_buffer. This was originally
designed for handling future IPs which could trigger a buffer overflow
interrupt. This patch gets rid of the reset_buffer callback altogether
and performs the actions in update_buffer, making it return the size
collected. We can always add the support for handling the overflow
interrupt case later.

This removes some not-so pretty hack (storing the new head in the
size field for snapshot mode) and cleans it up a little bit.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Convert driver messages to dev_dbg
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:53 +0000 (13:17 -0600)]
coresight: Convert driver messages to dev_dbg

Convert component enable/disable messages from dev_info to dev_dbg.
When used with perf, the components in the paths are enabled/disabled
during each schedule of the run, which can flood the dmesg with these
messages. Moreover, they are only useful for debug purposes. So,
convert such messages to dev_dbg() which can be turned on as
needed.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc-etr: Relax collection of trace from sysfs mode
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:52 +0000 (13:17 -0600)]
coresight: tmc-etr: Relax collection of trace from sysfs mode

Since the ETR now uses mode specific buffers, we can reliably
provide the trace data captured in sysfs mode, even when the ETR
is operating in PERF mode.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: tmc-etr: Handle driver mode specific ETR buffers
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:51 +0000 (13:17 -0600)]
coresight: tmc-etr: Handle driver mode specific ETR buffers

Since the ETR could be driven either by SYSFS or by perf, it
becomes complicated how we deal with the buffers used for each
of these modes. The ETR driver cannot simply free the current
attached buffer without knowing the provider (i.e, sysfs vs perf).

To solve this issue, we provide:
1) the driver-mode specific etr buffer to be retained in the drvdata
2) the etr_buf for a session should be passed on when enabling the
   hardware, which will be stored in drvdata->etr_buf. This will be
   replaced (not free'd) as soon as the hardware is disabled, after
   necessary sync operation.

The advantages of this are :

1) The common code path doesn't need to worry about how to dispose
   an existing buffer, if it is about to start a new session with a
   different buffer, possibly in a different mode.
2) The driver mode can control its buffers and can get access to the
   saved session even when the hardware is operating in a different
   mode. (e.g, we can still access a trace buffer from a sysfs mode
   even if the etr is now used in perf mode, without disrupting the
   current session.)

Towards this, we introduce a sysfs specific data which will hold the
etr_buf used for sysfs mode of operation, controlled solely by the
sysfs mode handling code.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Disable trace path upon source error
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:50 +0000 (13:17 -0600)]
coresight: perf: Disable trace path upon source error

We enable the trace path, before activating the source.
If we fail to enable the source, we must disable the path
to make sure it is available for another session.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Allow tracing on hotplugged CPUs
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:49 +0000 (13:17 -0600)]
coresight: perf: Allow tracing on hotplugged CPUs

At the moment, if there is no CPU specified for a given
event, we use cpu_online_mask and try to build path for
each of the CPUs in the mask. This could prevent any CPU
that is turned online later to be used for the tracing.

This patch changes to use the cpu_present_mask and tries
to build path for as much CPUs as possible ignoring the
failures in building path for some of the CPUs. If ever
we try to trace on those CPUs, we fail the operation.

Based on a patch from Mathieu Poirier.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Avoid unncessary CPU hotplug read lock
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:48 +0000 (13:17 -0600)]
coresight: perf: Avoid unncessary CPU hotplug read lock

We hold the read lock on CPU hotplug to simply copy the
online mask, which is not really needed. And this can
cause a lockdep warning, like :

[   54.632093] ======================================================
[   54.638207] WARNING: possible circular locking dependency detected
[   54.644322] 4.18.0-rc3-00042-g2d39e6356bb7-dirty #309 Not tainted
[   54.650350] ------------------------------------------------------
[   54.656464] perf/2862 is trying to acquire lock:
[   54.661031] 000000007e21d170 (&event->mmap_mutex){+.+.}, at: perf_event_set_output+0x98/0x138
[   54.669486]
[   54.669486] but task is already holding lock:
[   54.675256] 000000001080eb1b (&cpuctx_mutex){+.+.}, at: perf_event_ctx_lock_nested+0xf8/0x1f0
[   54.683704]
[   54.683704] which lock already depends on the new lock.
[   54.683704]
[   54.691797]
[   54.691797] the existing dependency chain (in reverse order) is:
[   54.699201]
[   54.699201] -> #3 (&cpuctx_mutex){+.+.}:
[   54.704556]        __mutex_lock+0x70/0x808
[   54.708608]        mutex_lock_nested+0x1c/0x28
[   54.713005]        perf_event_init_cpu+0x8c/0xd8
[   54.717574]        perf_event_init+0x194/0x1d4
[   54.721971]        start_kernel+0x2b8/0x42c
[   54.726107]
[   54.726107] -> #2 (pmus_lock){+.+.}:
[   54.731114]        __mutex_lock+0x70/0x808
[   54.735165]        mutex_lock_nested+0x1c/0x28
[   54.739560]        perf_event_init_cpu+0x30/0xd8
[   54.744129]        cpuhp_invoke_callback+0x84/0x248
[   54.748954]        _cpu_up+0xe8/0x1c8
[   54.752576]        do_cpu_up+0xa8/0xc8
[   54.756283]        cpu_up+0x10/0x18
[   54.759731]        smp_init+0xa0/0x114
[   54.763438]        kernel_init_freeable+0x120/0x288
[   54.768264]        kernel_init+0x10/0x108
[   54.772230]        ret_from_fork+0x10/0x18
[   54.776279]
[   54.776279] -> #1 (cpu_hotplug_lock.rw_sem){++++}:
[   54.782492]        cpus_read_lock+0x34/0xb0
[   54.786631]        etm_setup_aux+0x5c/0x308
[   54.790769]        rb_alloc_aux+0x1ec/0x300
[   54.794906]        perf_mmap+0x284/0x610
[   54.798787]        mmap_region+0x388/0x570
[   54.802838]        do_mmap+0x344/0x4f8
[   54.806544]        vm_mmap_pgoff+0xe4/0x110
[   54.810682]        ksys_mmap_pgoff+0xa8/0x240
[   54.814992]        sys_mmap+0x18/0x28
[   54.818613]        el0_svc_naked+0x30/0x34
[   54.822661]
[   54.822661] -> #0 (&event->mmap_mutex){+.+.}:
[   54.828445]        lock_acquire+0x48/0x68
[   54.832409]        __mutex_lock+0x70/0x808
[   54.836459]        mutex_lock_nested+0x1c/0x28
[   54.840855]        perf_event_set_output+0x98/0x138
[   54.845680]        _perf_ioctl+0x2a0/0x6a0
[   54.849731]        perf_ioctl+0x3c/0x68
[   54.853526]        do_vfs_ioctl+0xb8/0xa20
[   54.857577]        ksys_ioctl+0x80/0xb8
[   54.861370]        sys_ioctl+0xc/0x18
[   54.864990]        el0_svc_naked+0x30/0x34
[   54.869039]
[   54.869039] other info that might help us debug this:
[   54.869039]
[   54.876960] Chain exists of:
[   54.876960]   &event->mmap_mutex --> pmus_lock --> &cpuctx_mutex
[   54.876960]
[   54.887217]  Possible unsafe locking scenario:
[   54.887217]
[   54.893073]        CPU0                    CPU1
[   54.897552]        ----                    ----
[   54.902030]   lock(&cpuctx_mutex);
[   54.905396]                                lock(pmus_lock);
[   54.910911]                                lock(&cpuctx_mutex);
[   54.916770]   lock(&event->mmap_mutex);
[   54.920566]
[   54.920566]  *** DEADLOCK ***
[   54.920566]
[   54.926424] 1 lock held by perf/2862:
[   54.930042]  #0: 000000001080eb1b (&cpuctx_mutex){+.+.}, at: perf_event_ctx_lock_nested+0xf8/0x1f0

Since we have per-cpu array for the paths, we simply don't care about
the number of online CPUs. This patch gets rid of the
{get/put}_online_cpus().

Reported-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: perf: Fix per cpu path management
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:47 +0000 (13:17 -0600)]
coresight: perf: Fix per cpu path management

We create a coresight trace path for each online CPU when
we start the event. We rely on the number of online CPUs
and then go on to allocate an array matching the "number of
online CPUs" for holding the path and then uses normal
CPU id as the index to the array. This is problematic as
we could have some offline CPUs causing us to access beyond
the actual array size (e.g, on a dual SMP system, if CPU0 is
offline, CPU1 could be really accessing beyond the array).
The solution is to switch to per-cpu array for holding the path.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: etb10: Fix handling of perf mode
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:46 +0000 (13:17 -0600)]
coresight: etb10: Fix handling of perf mode

If the ETB is already enabled in sysfs mode, the ETB reports
success even if a perf mode is requested. Fix this by checking
the requested mode.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Fix handling of sinks
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:45 +0000 (13:17 -0600)]
coresight: Fix handling of sinks

The coresight components could be operated either in sysfs mode or in perf
mode. For some of the components, the mode of operation doesn't matter as
they simply relay the data to the next component in the trace path. But for
sinks, they need to be able to provide the trace data back to the user.
Thus we need to make sure that "mode" is handled appropriately. e.g,
the sysfs mode could have multiple sources driving the trace data, while
perf mode doesn't allow sharing the sink.

The coresight_enable_sink() however doesn't really allow this check to
trigger as it skips the "enable_sink" callback if the component is
already enabled, irrespective of the mode. This could cause mixing
of data from different modes or even same mode (in perf), if the
sources are different. Also, if we fail to enable the sink while
enabling a path (where sink is the first component enabled),
we could end up in disabling the components in the "entire"
path which were not enabled in this trial, causing disruptions
in the existing trace paths.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Use ERR_CAST instead of ERR_PTR
zhong jiang [Thu, 20 Sep 2018 19:17:44 +0000 (13:17 -0600)]
coresight: Use ERR_CAST instead of ERR_PTR

Use ERR_CAT inlined function to replace the ERR_PTR(PTR_ERR). It
make the code more concise.

Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Cleanup coresight DT bindings
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:43 +0000 (13:17 -0600)]
coresight: Cleanup coresight DT bindings

The coresight drivers relied on default bindings for graph
in DT, while reusing the "reg" field of the "ports" to indicate
the actual hardware port number for the connections. This can
cause duplicate ports with same addresses, but different
direction. However, with the rules getting stricter for the
address mismatch with the label, it is no longer possible to use
the port address field for the hardware port number.

This patch introduces new DT binding rules for coresight
components, based on the same generic DT graph bindings, but
avoiding the address duplication.

- All output ports must be specified under a child node with
  name "out-ports".
- All input ports must be specified under a childe node with
  name "in-ports".
- Port address should match the hardware port number.

The support for legacy bindings is retained, with a warning.

Cc: Sudeep Holla <sudeep.holla@arm.com>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: platform: Cleanup coresight connection handling
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:42 +0000 (13:17 -0600)]
coresight: platform: Cleanup coresight connection handling

The platform code parses the component connections and populates
a platform-description of the output connections in arrays of fields
(which is never freed). This is later copied in the coresight_register
to a newly allocated area, represented by coresight_connection(s).

This patch cleans up the code dealing with connections by making
use of the "coresight_connection" structure right at the platform
code and lets the generic driver simply re-use information provided
by the platform.

Thus making it reader friendly as well as avoiding the wastage of
unused memory.

Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Add helper to check if the endpoint is input
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:41 +0000 (13:17 -0600)]
coresight: Add helper to check if the endpoint is input

Add a helper to check if the given endpoint is input.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Fix remote endpoint parsing
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:40 +0000 (13:17 -0600)]
coresight: Fix remote endpoint parsing

When parsing the remote endpoint of an output port, we do :
     rport = of_graph_get_remote_port(ep);
     rparent = of_graph_get_remote_port_parent(ep);

and then parse the "remote_port" as if it was the remote endpoint,
which is wrong. The code worked fine because we used endpoint number
as the port number. Let us fix it and optimise a bit as:

     remote_ep = of_graph_get_remote_endpoint(ep);
     if (remote_ep)
        remote_parent = of_graph_get_port_parent(remote_ep);

and then, parse the remote_ep for the port/endpoint details.

Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: platform: Fix leaking device reference
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:39 +0000 (13:17 -0600)]
coresight: platform: Fix leaking device reference

We don't drop the reference on the remote device while parsing the
connection, held by bus_find_device(). Fix this by duplicating the
device name and dropping the reference.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Kim Phillips <kim.phillips@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: platform: Fix refcounting for graph nodes
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:38 +0000 (13:17 -0600)]
coresight: platform: Fix refcounting for graph nodes

The coresight driver doesn't drop the references on the
remote endpoint/port nodes. Add the missing of_node_put()
calls.

Reported-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: platform: Refactor graph endpoint parsing
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:37 +0000 (13:17 -0600)]
coresight: platform: Refactor graph endpoint parsing

Refactor the of graph endpoint parsing code, to make the error
handling easier.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agocoresight: Document error handling in coresight_register
Suzuki K Poulose [Thu, 20 Sep 2018 19:17:36 +0000 (13:17 -0600)]
coresight: Document error handling in coresight_register

commit 6403587a930c ("coresight: use put_device() instead of kfree()")
fixes the double freeing of resources and ensures that the device
refcount is dropped properly. Add a comment to explain this to
help the readers and prevent people trying to "unfix" it again.

While at it, rename the labels for better readability.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agomisc: sgi-gru: fix fall-through annotations
Gustavo A. R. Silva [Thu, 13 Sep 2018 18:53:18 +0000 (13:53 -0500)]
misc: sgi-gru: fix fall-through annotations

Replace "fallthru" with a proper "fall through" annotation.

This fix is part of the ongoing efforts to enabling
-Wimplicit-fallthrough

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Dimitri Sivanich <sivanich@hpe.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoMerge b4.19-rc4 into char-misc-next
Greg Kroah-Hartman [Sun, 16 Sep 2018 20:41:47 +0000 (22:41 +0200)]
Merge b4.19-rc4 into char-misc-next

We want the bugfixes in here as well.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoLinux 4.19-rc4
Linus Torvalds [Sun, 16 Sep 2018 18:52:37 +0000 (11:52 -0700)]
Linux 4.19-rc4

6 years agoCode of Conduct: Let's revamp it.
Greg Kroah-Hartman [Sat, 15 Sep 2018 18:26:44 +0000 (20:26 +0200)]
Code of Conduct: Let's revamp it.

The Code of Conflict is not achieving its implicit goal of fostering
civility and the spirit of 'be excellent to each other'.  Explicit
guidelines have demonstrated success in other projects and other areas
of the kernel.

Here is a Code of Conduct statement for the wider kernel.  It is based
on the Contributor Covenant as described at www.contributor-covenant.org

From this point forward, we should abide by these rules in order to help
make the kernel community a welcoming environment to participate in.

Signed-off-by: Chris Mason <clm@fb.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Olof Johansson <olof@lxom.net>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 15 Sep 2018 18:02:46 +0000 (08:02 -1000)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingol Molnar:
 "Misc fixes:

   - EFI crash fix

   - Xen PV fixes

   - do not allow PTI on 2-level 32-bit kernels for now

   - documentation fix"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/APM: Fix build warning when PROC_FS is not enabled
  Revert "x86/mm/legacy: Populate the user page-table with user pgd's"
  x86/efi: Load fixmap GDT in efi_call_phys_epilog() before setting %cr3
  x86/xen: Disable CPU0 hotplug for Xen PV
  x86/EISA: Don't probe EISA bus for Xen PV guests
  x86/doc: Fix Documentation/x86/earlyprintk.txt

6 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 15 Sep 2018 18:00:37 +0000 (08:00 -1000)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Misc fixes: various scheduler metrics corner case fixes, a
  sched_features deadlock fix, and a topology fix for certain NUMA
  systems"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/fair: Fix kernel-doc notation warning
  sched/fair: Fix load_balance redo for !imbalance
  sched/fair: Fix scale_rt_capacity() for SMT
  sched/fair: Fix vruntime_normalized() for remote non-migration wakeup
  sched/pelt: Fix update_blocked_averages() for RT and DL classes
  sched/topology: Set correct NUMA topology type
  sched/debug: Fix potential deadlock when writing to sched_features

6 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 15 Sep 2018 16:44:32 +0000 (06:44 -1000)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Mostly tooling fixes, but also breakpoint and x86 PMU driver fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (23 commits)
  perf tools: Fix maps__find_symbol_by_name()
  tools headers uapi: Update tools's copy of linux/if_link.h
  tools headers uapi: Update tools's copy of linux/vhost.h
  tools headers uapi: Update tools's copies of kvm headers
  tools headers uapi: Update tools's copy of drm/drm.h
  tools headers uapi: Update tools's copy of asm-generic/unistd.h
  tools headers uapi: Update tools's copy of linux/perf_event.h
  perf/core: Force USER_DS when recording user stack data
  perf/UAPI: Clearly mark __PERF_SAMPLE_CALLCHAIN_EARLY as internal use
  perf/x86/intel: Add support/quirk for the MISPREDICT bit on Knights Landing CPUs
  perf annotate: Fix parsing aarch64 branch instructions after objdump update
  perf probe powerpc: Ignore SyS symbols irrespective of endianness
  perf event-parse: Use fixed size string for comms
  perf util: Fix bad memory access in trace info.
  perf tools: Streamline bpf examples and headers installation
  perf evsel: Fix potential null pointer dereference in perf_evsel__new_idx()
  perf arm64: Fix include path for asm-generic/unistd.h
  perf/hw_breakpoint: Simplify breakpoint enable in perf_event_modify_breakpoint
  perf/hw_breakpoint: Enable breakpoint in modify_user_hw_breakpoint
  perf/hw_breakpoint: Remove superfluous bp->attr.disabled = 0
  ...

6 years agoMerge branch 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 15 Sep 2018 16:39:09 +0000 (06:39 -1000)]
Merge branch 'locking-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull locking fixes from Ingo Molnar:
 "Misc fixes: liblockdep fixes and ww_mutex fixes"

* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/ww_mutex: Fix spelling mistake "cylic" -> "cyclic"
  locking/lockdep: Delete unnecessary #include
  tools/lib/lockdep: Add dummy task_struct state member
  tools/lib/lockdep: Add empty nmi.h
  tools/lib/lockdep: Update Sasha Levin email to MSFT
  jump_label: Fix typo in warning message
  locking/mutex: Fix mutex debug call and ww_mutex documentation

6 years agox86/APM: Fix build warning when PROC_FS is not enabled
Randy Dunlap [Fri, 14 Sep 2018 22:10:29 +0000 (15:10 -0700)]
x86/APM: Fix build warning when PROC_FS is not enabled

Fix build warning in apm_32.c when CONFIG_PROC_FS is not enabled:

../arch/x86/kernel/apm_32.c:1643:12: warning: 'proc_apm_show' defined but not used [-Wunused-function]
 static int proc_apm_show(struct seq_file *m, void *v)

Fixes: 3f3942aca6da ("proc: introduce proc_create_single{,_data}")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Jiri Kosina <jikos@kernel.org>
Link: https://lkml.kernel.org/r/be39ac12-44c2-4715-247f-4dcc3c525b8b@infradead.org
6 years agoMerge tag '4.19-rc3-smb3-cifs' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 15 Sep 2018 05:33:42 +0000 (19:33 -1000)]
Merge tag '4.19-rc3-smb3-cifs' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Fixes for four CIFS/SMB3 potential pointer overflow issues, one minor
  build fix, and a build warning cleanup"

* tag '4.19-rc3-smb3-cifs' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: read overflow in is_valid_oplock_break()
  cifs: integer overflow in in SMB2_ioctl()
  CIFS: fix wrapping bugs in num_entries()
  cifs: prevent integer overflow in nxt_dir_entry()
  fs/cifs: require sha512
  fs/cifs: suppress a string overflow warning

6 years agoMerge tag 'nfs-for-4.19-2' of git://git.linux-nfs.org/projects/anna/linux-nfs
Linus Torvalds [Sat, 15 Sep 2018 05:25:28 +0000 (19:25 -1000)]
Merge tag 'nfs-for-4.19-2' of git://git.linux-nfs.org/projects/anna/linux-nfs

Pull NFS client bugfixes from Anna Schumaker:
 "These are a handful of fixes for problems that Trond found. Patch #1
  and #3 have the same name, a second issue was found after applying the
  first patch.

  Stable bugfixes:
   - v4.17+: Fix tracepoint Oops in initiate_file_draining()
   - v4.11+: Fix an infinite loop on I/O

  Other fixes:
   - Return errors if a waiting layoutget is killed
   - Don't open code clearing of delegation state"

* tag 'nfs-for-4.19-2' of git://git.linux-nfs.org/projects/anna/linux-nfs:
  NFS: Don't open code clearing of delegation state
  NFSv4.1 fix infinite loop on I/O.
  NFSv4: Fix a tracepoint Oops in initiate_file_draining()
  pNFS: Ensure we return the error if someone kills a waiting layoutget
  NFSv4: Fix a tracepoint Oops in initiate_file_draining()

6 years agoMerge tag 'trace-v4.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Sat, 15 Sep 2018 03:31:02 +0000 (17:31 -1000)]
Merge tag 'trace-v4.19-rc3' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "This fixes an issue with the build system caused by a change that
  modifies CC_FLAGS_FTRACE. The issue is that it breaks the dependencies
  and causes "make targz-pkg" to rebuild the entire world"

* tag 'trace-v4.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing/Makefile: Fix handling redefinition of CC_FLAGS_FTRACE

6 years agoMerge tag 'devicetree-fixes-for-4.19-2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 14 Sep 2018 23:03:17 +0000 (13:03 -1000)]
Merge tag 'devicetree-fixes-for-4.19-2' of git://git./linux/kernel/git/robh/linux

Pull DeviceTree fix from Rob Herring:
 "One regression for a 20 year old PowerMac:

   - Fix a regression on systems having a DT without any phandles which
     happens on a PowerMac G3"

* tag 'devicetree-fixes-for-4.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  of: fix phandle cache creation for DTs with no phandles

6 years agoMerge tag 'for-linus-4.19c-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 14 Sep 2018 23:01:06 +0000 (13:01 -1000)]
Merge tag 'for-linus-4.19c-rc4-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:
 "This contains some minor cleanups and fixes:

   - a new knob for controlling scrubbing of pages returned by the Xen
     balloon driver to the Xen hypervisor to address a boot performance
     issue seen in large guests booted pre-ballooned

   - a fix of a regression in the gntdev driver which made it impossible
     to use fully virtualized guests (HVM guests) with a 4.19 based dom0

   - a fix in Xen cpu hotplug functionality which could be triggered by
     wrong admin commands (setting number of active vcpus to 0)

  One further note: the patches have all been under test for several
  days in another branch. This branch has been rebased in order to avoid
  merge conflicts"

* tag 'for-linus-4.19c-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/gntdev: fix up blockable calls to mn_invl_range_start
  xen: fix GCC warning and remove duplicate EVTCHN_ROW/EVTCHN_COL usage
  xen: avoid crash in disable_hotplug_cpu
  xen/balloon: add runtime control for scrubbing ballooned out pages
  xen/manage: don't complain about an empty value in control/sysrq node

6 years agoMerge tag 'xtensa-20180914' of git://github.com/jcmvbkbc/linux-xtensa
Linus Torvalds [Fri, 14 Sep 2018 22:56:42 +0000 (12:56 -1000)]
Merge tag 'xtensa-20180914' of git://github.com/jcmvbkbc/linux-xtensa

Pull Xtensa fixes and cleanups from Max Filippov:

 - don't allocate memory in platform_setup as the memory allocator is
   not initialized at that point yet;

 - remove unnecessary ifeq KBUILD_SRC from arch/xtensa/Makefile;

 - enable SG chaining in arch/xtensa/Kconfig.

* tag 'xtensa-20180914' of git://github.com/jcmvbkbc/linux-xtensa:
  xtensa: enable SG chaining in Kconfig
  xtensa: remove unnecessary KBUILD_SRC ifeq conditional
  xtensa: ISS: don't allocate memory in platform_setup

6 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 14 Sep 2018 22:42:02 +0000 (12:42 -1000)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "The trickle of arm64 fixes continues to come in.

  Nothing that's the end of the world, but we've got a fix for PCI IO
  port accesses, an accidental naked "asm goto" and a fix to the
  vmcoreinfo PT_NOTE merged this time around which we'd like to get
  sorted before it becomes ABI.

   - Fix ioport_map() mapping the wrong physical address for some I/O
     BARs

   - Remove direct use of "asm goto", since some compilers don't like
     that

   - Ensure kimage_voffset is always present in vmcoreinfo PT_NOTE"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  asm-generic: io: Fix ioport_map() for !CONFIG_GENERIC_IOMAP && CONFIG_INDIRECT_PIO
  arm64: kernel: arch_crash_save_vmcoreinfo() should depend on CONFIG_CRASH_CORE
  arm64: jump_label.h: use asm_volatile_goto macro instead of "asm goto"

6 years agoNFS: Don't open code clearing of delegation state
Trond Myklebust [Wed, 5 Sep 2018 18:07:15 +0000 (14:07 -0400)]
NFS: Don't open code clearing of delegation state

Add a helper for the case when the nfs4 open state has been set to use
a delegation stateid, and we want to revert to using the open stateid.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
6 years agoNFSv4.1 fix infinite loop on I/O.
Trond Myklebust [Wed, 5 Sep 2018 18:07:14 +0000 (14:07 -0400)]
NFSv4.1 fix infinite loop on I/O.

The previous fix broke recovery of delegated stateids because it assumes
that if we did not mark the delegation as suspect, then the delegation has
effectively been revoked, and so it removes that delegation irrespectively
of whether or not it is valid and still in use. While this is "mostly
harmless" for ordinary I/O, we've seen pNFS fail with LAYOUTGET spinning
in an infinite loop while complaining that we're using an invalid stateid
(in this case the all-zero stateid).

What we rather want to do here is ensure that the delegation is always
correctly marked as needing testing when that is the case. So we want
to close the loophole offered by nfs4_schedule_stateid_recovery(),
which marks the state as needing to be reclaimed, but not the
delegation that may be backing it.

Fixes: 0e3d3e5df07dc ("NFSv4.1 fix infinite loop on IO BAD_STATEID error")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: stable@vger.kernel.org # v4.11+
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
6 years agoNFSv4: Fix a tracepoint Oops in initiate_file_draining()
Trond Myklebust [Wed, 5 Sep 2018 18:07:13 +0000 (14:07 -0400)]
NFSv4: Fix a tracepoint Oops in initiate_file_draining()

Now that the value of 'ino' can be NULL or an ERR_PTR(), we need to
change the test in the tracepoint.

Fixes: ce5624f7e6675 ("NFSv4: Return NFS4ERR_DELAY when a layout fails...")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: stable@vger.kernel.org # v4.17+
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
6 years agopNFS: Ensure we return the error if someone kills a waiting layoutget
Trond Myklebust [Wed, 5 Sep 2018 18:07:12 +0000 (14:07 -0400)]
pNFS: Ensure we return the error if someone kills a waiting layoutget

If someone interrupts a wait on one or more outstanding layoutgets in
pnfs_update_layout() then return the ERESTARTSYS/EINTR error.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
6 years agoNFSv4: Fix a tracepoint Oops in initiate_file_draining()
Trond Myklebust [Thu, 23 Aug 2018 15:02:49 +0000 (11:02 -0400)]
NFSv4: Fix a tracepoint Oops in initiate_file_draining()

Now that the value of 'ino' can be NULL or an ERR_PTR(), we need to
change the test in the tracepoint.

Fixes: ce5624f7e6675 ("NFSv4: Return NFS4ERR_DELAY when a layout fails...")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: stable@vger.kernel.org # v4.17+
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
6 years agoMerge tag 'dmaengine-fix-4.19-rc4' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Fri, 14 Sep 2018 16:08:40 +0000 (06:08 -1000)]
Merge tag 'dmaengine-fix-4.19-rc4' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fix from Vinod Koul:
 "Fix the mic_x100_dma driver to use devm_kzalloc for driver memory, so
  that it is freed properly when it unregisters from dmaengine using
  managed API"

* tag 'dmaengine-fix-4.19-rc4' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: mic_x100_dma: use devm_kzalloc to fix an issue

6 years agoMerge tag 'usb-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Fri, 14 Sep 2018 15:59:48 +0000 (05:59 -1000)]
Merge tag 'usb-4.19-rc4' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are a number of small USB driver fixes for -rc4.

  The usual suspects of gadget, xhci, and dwc2/3 are in here, along with
  some reverts of reported problem changes, and a number of build
  documentation warning fixes. Full details are in the shortlog.

  All of these have been in linux-next with no reported issues"

* tag 'usb-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (28 commits)
  Revert "cdc-acm: implement put_char() and flush_chars()"
  usb: Change usb_of_get_companion_dev() place to usb/common
  usb: xhci: fix interrupt transfer error happened on MTK platforms
  usb: cdc-wdm: Fix a sleep-in-atomic-context bug in service_outstanding_interrupt()
  usb: misc: uss720: Fix two sleep-in-atomic-context bugs
  usb: host: u132-hcd: Fix a sleep-in-atomic-context bug in u132_get_frame()
  usb: Avoid use-after-free by flushing endpoints early in usb_set_interface()
  linux/mod_devicetable.h: fix kernel-doc missing notation for typec_device_id
  usb/typec: fix kernel-doc notation warning for typec_match_altmode
  usb: Don't die twice if PCI xhci host is not responding in resume
  usb: mtu3: fix error of xhci port id when enable U3 dual role
  usb: uas: add support for more quirk flags
  USB: Add quirk to support DJI CineSSD
  usb: typec: fix kernel-doc parameter warning
  usb/dwc3/gadget: fix kernel-doc parameter warning
  USB: yurex: Check for truncation in yurex_read()
  USB: yurex: Fix buffer over-read in yurex_write()
  usb: host: xhci-plat: Iterate over parent nodes for finding quirks
  xhci: Fix use after free for URB cancellation on a reallocated endpoint
  USB: add quirk for WORLDE Controller KS49 or Prodipe MIDI 49C USB controller
  ...

6 years agoMerge tag 'tty-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Fri, 14 Sep 2018 15:58:12 +0000 (05:58 -1000)]
Merge tag 'tty-4.19-rc4' of git://git./linux/kernel/git/gregkh/tty

Pull tty fixes from Greg KH:
 "Here are three small HVC tty driver fixes to resolve a reported
  regression from 4.19-rc1.

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'tty-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: hvc: hvc_write() fix break condition
  tty: hvc: hvc_poll() fix read loop batching
  tty: hvc: hvc_poll() fix read loop hang

6 years agoMerge tag 'staging-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Fri, 14 Sep 2018 15:43:43 +0000 (05:43 -1000)]
Merge tag 'staging-4.19-rc4' of git://git./linux/kernel/git/gregkh/staging

Pull staging/IIO driver fixes from Greg KH:
 "Here are a few small staging and iio driver fixes for -rc4.

  Nothing major, just a few small bugfixes for some reported issues, and
  a MAINTAINERS file update for the fbtft drivers.

  We also re-enable the building of the erofs filesystem as the XArray
  patches that were causing it to break never got merged in the -rc1
  cycle, so there's no reason it can't be turned back on for now. The
  problem that was previously there is now being handled in the Xarray
  tree at the moment, so it will not hit us again in the future.

  All of these patches have been in linux-next with no reported issues"

* tag 'staging-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: vboxvideo: Change address of scanout buffer on page-flip
  staging: vboxvideo: Fix IRQs no longer working
  staging: gasket: TODO: re-implement using UIO
  staging/fbtft: Update TODO and mailing lists
  staging: erofs: rename superblock flags (MS_xyz -> SB_xyz)
  iio: imu: st_lsm6dsx: take into account ts samples in wm configuration
  Revert "iio: temperature: maxim_thermocouple: add MAX31856 part"
  Revert "staging: erofs: disable compiling temporarile"
  MAINTAINERS: Switch a maintainer for drivers/staging/gasket
  staging: wilc1000: revert "fix TODO to compile spi and sdio components in single module"

6 years agoMerge tag 'char-misc-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Fri, 14 Sep 2018 15:40:13 +0000 (05:40 -1000)]
Merge tag 'char-misc-4.19-rc4' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are a small handful of char/misc driver fixes for 4.19-rc4.

  All of them are simple, resolving reported problems in a few drivers.
  Full details are in the shortlog.

  All of these have been in linux-next with no reported issues"

* tag 'char-misc-4.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  firmware: Fix security issue with request_firmware_into_buf()
  vmbus: don't return values for uninitalized channels
  fpga: dfl: fme: fix return value check in in pr_mgmt_init()
  misc: hmc6352: fix potential Spectre v1
  Tools: hv: Fix a bug in the key delete code
  misc: ibmvsm: Fix wrong assignment of return code
  android: binder: fix the race mmap and alloc_new_buf_locked
  mei: bus: need to unlink client before freeing
  mei: bus: fix hw module get/put balance
  mei: fix use-after-free in mei_cl_write
  mei: ignore not found client in the enumeration

6 years agoRevert "x86/mm/legacy: Populate the user page-table with user pgd's"
Joerg Roedel [Fri, 14 Sep 2018 10:59:14 +0000 (12:59 +0200)]
Revert "x86/mm/legacy: Populate the user page-table with user pgd's"

This reverts commit 1f40a46cf47c12d93a5ad9dccd82bd36ff8f956a.

It turned out that this patch is not sufficient to enable PTI on 32 bit
systems with legacy 2-level page-tables. In this paging mode the huge-page
PTEs are in the top-level page-table directory, where also the mirroring to
the user-space page-table happens. So every huge PTE exits twice, in the
kernel and in the user page-table.

That means that accessed/dirty bits need to be fetched from two PTEs in
this mode to be safe, but this is not trivial to implement because it needs
changes to generic code just for the sake of enabling PTI with 32-bit
legacy paging. As all systems that need PTI should support PAE anyway,
remove support for PTI when 32-bit legacy paging is used.

Fixes: 7757d607c6b3 ('x86/pti: Allow CONFIG_PAGE_TABLE_ISOLATION for x86_32')
Reported-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: hpa@zytor.com
Cc: linux-mm@kvack.org
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Link: https://lkml.kernel.org/r/1536922754-31379-1-git-send-email-joro@8bytes.org
6 years agofirmware: coreboot: Only populate devices in coreboot_table_init()
Stephen Boyd [Wed, 15 Aug 2018 20:37:08 +0000 (13:37 -0700)]
firmware: coreboot: Only populate devices in coreboot_table_init()

This function checks the header for sanity, registers a bus, and
populates devices for each coreboot table entry. Let's just populate
devices here and pull the other bits up into the caller so that this
function can be repurposed for pure device creation and registration.

Cc: Wei-Ning Huang <wnhuang@chromium.org>
Cc: Julius Werner <jwerner@chromium.org>
Cc: Brian Norris <briannorris@chromium.org>
Cc: Samuel Holland <samuel@sholland.org>
Suggested-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agofirmware: coreboot: Remap RAM with memremap() instead of ioremap()
Stephen Boyd [Wed, 15 Aug 2018 20:37:07 +0000 (13:37 -0700)]
firmware: coreboot: Remap RAM with memremap() instead of ioremap()

This is all system memory, so we shouldn't be mapping this all with
ioremap() as these aren't I/O regions. Instead, they're memory regions
so we should use memremap(). Pick MEMREMAP_WB so we can map memory from
RAM directly if that's possible, otherwise it falls back to
ioremap_cache() like is being done here already. This also nicely
silences the sparse warnings in this code and reduces the need to copy
anything around anymore.

Cc: Wei-Ning Huang <wnhuang@chromium.org>
Cc: Julius Werner <jwerner@chromium.org>
Cc: Brian Norris <briannorris@chromium.org>
Cc: Samuel Holland <samuel@sholland.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agofirmware: coreboot: Collapse platform drivers into bus core
Stephen Boyd [Wed, 15 Aug 2018 20:37:06 +0000 (13:37 -0700)]
firmware: coreboot: Collapse platform drivers into bus core

The DT based and ACPI based platform drivers here do the same thing; map
some memory and hand it over to the coreboot bus to populate devices.
The only major difference is that the DT based driver doesn't map the
coreboot table header to figure out how large of a region to map for the
whole coreboot table and it uses of_iomap() instead of ioremap_cache().
A cached or non-cached mapping shouldn't matter here and mapping some
smaller region first before mapping the whole table is just more work
but should be OK. In the end, we can remove two files and combine the
code all in one place making it easier to reason about things.

We leave the old Kconfigs in place for a little while longer but make
them hidden and select the previously hidden config option. This way
users can upgrade without having to know to reselect this config in the
future. Later on we can remove the old hidden configs.

Cc: Wei-Ning Huang <wnhuang@chromium.org>
Cc: Julius Werner <jwerner@chromium.org>
Cc: Brian Norris <briannorris@chromium.org>
Cc: Samuel Holland <samuel@sholland.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agofirmware: coreboot: Make bus registration symmetric
Stephen Boyd [Wed, 15 Aug 2018 20:37:05 +0000 (13:37 -0700)]
firmware: coreboot: Make bus registration symmetric

The bus is registered in module_init() but is unregistered when the
platform driver remove() function calls coreboot_table_exit(). That
isn't symmetric and it causes the bus to appear on systems that compile
this code in, even when there isn't any coreboot firmware on the device.
Let's move the registration to the coreboot_table_init() function so
that it matches the exit path.

Cc: Wei-Ning Huang <wnhuang@chromium.org>
Cc: Julius Werner <jwerner@chromium.org>
Cc: Brian Norris <briannorris@chromium.org>
Cc: Samuel Holland <samuel@sholland.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agofirmware: coreboot: Unmap ioregion after device population
Stephen Boyd [Wed, 15 Aug 2018 20:37:04 +0000 (13:37 -0700)]
firmware: coreboot: Unmap ioregion after device population

Both callers of coreboot_table_init() ioremap the pointer that comes in
but they don't unmap the memory on failure. Both of them also fail probe
immediately with the return value of coreboot_table_init(), leaking a
mapping when it fails. The mapping isn't necessary at all after devices
are populated either, so we can just drop the mapping here when we exit
the function. Let's do that to simplify the code a bit and plug the leak.

Cc: Wei-Ning Huang <wnhuang@chromium.org>
Cc: Julius Werner <jwerner@chromium.org>
Cc: Brian Norris <briannorris@chromium.org>
Cc: Samuel Holland <samuel@sholland.org>
Fixes: 570d30c2823f ("firmware: coreboot: Expose the coreboot table as a bus")
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agofirmware: coreboot: Let OF core populate platform device
Stephen Boyd [Wed, 15 Aug 2018 20:37:03 +0000 (13:37 -0700)]
firmware: coreboot: Let OF core populate platform device

Now that the /firmware/coreboot node in DT is populated by the core DT
platform code with commit 3aa0582fdb82 ("of: platform: populate
/firmware/ node from of_platform_default_populate_init()") we should and
can remove the platform device creation here. Otherwise, the
of_platform_device_create() call will fail, the coreboot of driver won't
be registered, and this driver will never bind. At the same time, we
should move this driver to use MODULE_DEVICE_TABLE so that module
auto-load works properly when the coreboot device is auto-populated and
we should drop the of_node handling that was presumably placed here to
hold a reference to the DT node created during module init that no
longer happens.

Cc: Wei-Ning Huang <wnhuang@chromium.org>
Cc: Julius Werner <jwerner@chromium.org>
Reviewed-by: Brian Norris <briannorris@chromium.org>
Cc: Samuel Holland <samuel@sholland.org>
Reviewed-by: Sudeep Holla <Sudeep.Holla@arm.com>
Fixes: 3aa0582fdb82 ("of: platform: populate /firmware/ node from of_platform_default_populate_init()")
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agodrivers: misc: ad525x_dpot: Update MODULE AUTHOR email address
Michael Hennerich [Tue, 14 Aug 2018 11:30:18 +0000 (13:30 +0200)]
drivers: misc: ad525x_dpot: Update MODULE AUTHOR email address

no functional changes

Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agomisc: sgi-xp: remove meaningless null check before kfree
zhong jiang [Fri, 17 Aug 2018 03:42:35 +0000 (11:42 +0800)]
misc: sgi-xp: remove meaningless null check before kfree

kfree has taken null pointer into account. so check the null pointer
before kfree is meaningless.

Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Acked-by: Robin Holt <robinmholt@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoslimbus: ngd: mark PM functions as __maybe_unused
Arnd Bergmann [Mon, 20 Aug 2018 09:35:37 +0000 (11:35 +0200)]
slimbus: ngd: mark PM functions as __maybe_unused

qcom_slim_ngd_runtime_suspend is protected by an #ifdef,
qcom_slim_ngd_runtime_idle is now, which causes a build time warning:

drivers/slimbus/qcom-ngd-ctrl.c:1470:12: error: 'qcom_slim_ngd_runtime_idle' defined but not used [-Werror=unused-function]

Marking both as __maybe_unused lets us get rid of the warning
as well as the #ifdef.

Fixes: 917809e2280b ("slimbus: ngd: Add qcom SLIMBus NGD driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Replace "x==NULL" to "!x"
Roman Kiryanov [Mon, 27 Aug 2018 18:23:15 +0000 (11:23 -0700)]
platform: goldfish: pipe: Replace "x==NULL" to "!x"

checkpatch: Comparison to NULL could be written "!x"

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Move logical ops to the end of the prev line
Roman Kiryanov [Mon, 27 Aug 2018 18:23:14 +0000 (11:23 -0700)]
platform: goldfish: pipe: Move logical ops to the end of the prev line

checkpatch: Logical continuations should be on the previous line

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Fix alignment to match parenthesis
Roman Kiryanov [Mon, 27 Aug 2018 18:23:13 +0000 (11:23 -0700)]
platform: goldfish: pipe: Fix alignment to match parenthesis

checkpatch: Alignment should match open parenthesis.

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Fix lines to not end an opening bracket
Roman Kiryanov [Mon, 27 Aug 2018 18:23:12 +0000 (11:23 -0700)]
platform: goldfish: pipe: Fix lines to not end an opening bracket

checkpatch: Lines should not end with a '(' or '['

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Rename goldfish_x to goldfish_pipe_x
Roman Kiryanov [Mon, 27 Aug 2018 18:23:11 +0000 (11:23 -0700)]
platform: goldfish: pipe: Rename goldfish_x to goldfish_pipe_x

Add "pipe" to the pipe related function names.

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Remove reduntant casting to (void)
Roman Kiryanov [Mon, 27 Aug 2018 18:23:10 +0000 (11:23 -0700)]
platform: goldfish: pipe: Remove reduntant casting to (void)

Casting to (void) is no-op.

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Rename 'wakeBit' to 'wake_bit'
Roman Kiryanov [Mon, 27 Aug 2018 18:23:09 +0000 (11:23 -0700)]
platform: goldfish: pipe: Rename 'wakeBit' to 'wake_bit'

checkpatch: Avoid CamelCase

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Remove a blank line before '}'
Roman Kiryanov [Mon, 27 Aug 2018 18:23:08 +0000 (11:23 -0700)]
platform: goldfish: pipe: Remove a blank line before '}'

checkpatch: Blank lines aren't necessary before a close brace '}'

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Remove reduntant casting
Roman Kiryanov [Mon, 27 Aug 2018 18:23:07 +0000 (11:23 -0700)]
platform: goldfish: pipe: Remove reduntant casting

Casting to u32 is not required here.

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
6 years agoplatform: goldfish: pipe: Replace two code blocks with a function call
Roman Kiryanov [Mon, 27 Aug 2018 18:23:06 +0000 (11:23 -0700)]
platform: goldfish: pipe: Replace two code blocks with a function call

Two function calls look cleaner because the function introduces
takes case of all bit shifting and casting.

Signed-off-by: Roman Kiryanov <rkir@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>