diff mbox

[01/11,v6] coresight: add CoreSight core layer framework

Message ID 1410450558-12358-2-git-send-email-mathieu.poirier@linaro.org
State New
Headers show

Commit Message

Mathieu Poirier Sept. 11, 2014, 3:49 p.m. UTC
From: Pratik Patel <pratikp@codeaurora.org>

CoreSight components are compliant with the ARM CoreSight
architecture specification and can be connected in various
topologies to suit a particular SoC tracing needs. These trace
components can generally be classified as sources, links and
sinks. Trace data produced by one or more sources flows through
the intermediate links connecting the source to the currently
selected sink.

The CoreSight framework provides an interface for the CoreSight trace
drivers to register themselves with. It's intended to build up a
topological view of the CoreSight components and configure the
correct serie of components on user input via debugfs.

For eg., when enabling a source, the framework builds up a path
consisting of all the components connecting the source to the
currently selected sink(s) and enables all of them.

The framework also supports switching between available sinks
and provides status information to user space applications
through the debugfs interface.

Signed-off-by: Pratik Patel <pratikp@codeaurora.org>
Signed-off-by: Panchaxari Prasannamurthy <panchaxari.prasannamurthy@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
---
 arch/arm/Kconfig.debug             |   9 +
 drivers/Makefile                   |   1 +
 drivers/amba/bus.c                 |   2 +-
 drivers/coresight/Makefile         |   5 +
 drivers/coresight/coresight-priv.h |  63 ++++
 drivers/coresight/coresight.c      | 663 +++++++++++++++++++++++++++++++++++++
 drivers/coresight/of_coresight.c   | 201 +++++++++++
 include/linux/amba/bus.h           |   1 +
 include/linux/coresight.h          | 275 +++++++++++++++
 9 files changed, 1219 insertions(+), 1 deletion(-)
 create mode 100644 drivers/coresight/Makefile
 create mode 100644 drivers/coresight/coresight-priv.h
 create mode 100644 drivers/coresight/coresight.c
 create mode 100644 drivers/coresight/of_coresight.c
 create mode 100644 include/linux/coresight.h

Comments

Greg Kroah-Hartman Sept. 11, 2014, 8:33 p.m. UTC | #1
Some first impressions in glancing at the code, not a complete review at
all:

On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
> --- /dev/null
> +++ b/drivers/coresight/coresight.c
> @@ -0,0 +1,663 @@
> +/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 and
> + * only version 2 as published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + */
> +
> +#define pr_fmt(fmt) "coresight: " fmt

MODULE_NAME ?

> +
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/init.h>
> +#include <linux/types.h>
> +#include <linux/device.h>
> +#include <linux/io.h>
> +#include <linux/err.h>
> +#include <linux/export.h>
> +#include <linux/slab.h>
> +#include <linux/semaphore.h>
> +#include <linux/clk.h>
> +#include <linux/coresight.h>
> +#include <linux/of_platform.h>
> +#include <linux/debugfs.h>
> +#include <linux/delay.h>
> +
> +#include "coresight-priv.h"
> +
> +struct dentry *cs_debugfs_parent = NULL;
> +
> +static LIST_HEAD(coresight_orph_conns);
> +static LIST_HEAD(coresight_devs);

You are a struct device, you don't need a separate list, why not just
iterate over your bus list of devices?

> +/**
> + * @id:		unique ID of the component.
> + * @conns:	list of connections associated to this component.
> + * @type:	as defined by @coresight_dev_type.
> + * @subtype:	as defined by @coresight_dev_subtype.
> + * @ops:	generic operations for this component, as defined
> +		by @coresight_ops.
> + * @de:		handle on a component's debugfs entry.
> + * @dev:	The device entity associated to this component.
> + * @kref:	keeping count on component references.
> + * @dev_link:	link of current component into list of all components
> +		discovered in the system.
> + * @path_link:	link of current component into the path being enabled.
> + * @owner:	typically "THIS_MODULE".
> + * @enable:	'true' if component is currently part of an active path.
> + * @activated:	'true' only if a _sink_ has been activated.  A sink can be
> +		activated but not yet enabled.  Enabling for a _sink_
> +		happens when a source has been selected for that it.
> + */
> +struct coresight_device {
> +	int id;

Why not use the device name instead?

> +	struct coresight_connection *conns;
> +	int nr_conns;
> +	enum coresight_dev_type type;
> +	struct coresight_dev_subtype subtype;
> +	const struct coresight_ops *ops;
> +	struct dentry *de;

All devices have a dentry?  in debugfs?  isn't that overkill?

> +	struct device dev;
> +	struct kref kref;

You CAN NOT have two reference counts in the same structure, that's a
huge design mistake.  Stick with one reference count, otherwise they are
guaranteed to get out of sync and bad things will happen.

> +	struct list_head dev_link;

As discussed above, I don't think you need this.

> +	struct list_head path_link;

Odds are, you don't need this either.

> +	struct module *owner;

devices aren't owned by modules, they are data, not code.


> +	bool enable;	/* true only if configured as part of a path */
> +	bool activated;	/* true only if a sink is part of a path */
> +};
> +
> +#define to_coresight_device(d) container_of(d, struct coresight_device, dev)
> +
> +#define source_ops(csdev)	csdev->ops->source_ops
> +#define sink_ops(csdev)		csdev->ops->sink_ops
> +#define link_ops(csdev)		csdev->ops->link_ops
> +
> +#define CORESIGHT_DEBUGFS_ENTRY(__name, __entry_name,			\
> +				 __mode, __get, __set, __fmt)		\
> +DEFINE_SIMPLE_ATTRIBUTE(__name ## _ops, __get, __set, __fmt)		\

Use the RW and RO only versions please.  No need to ever set your own
mode value.

thanks,

greg k-h
Mathieu Poirier Sept. 12, 2014, 5:41 p.m. UTC | #2
Good morning and thanks for the review.  Pls see comments below.

Mathieu

On 11 September 2014 14:33, Greg KH <gregkh@linuxfoundation.org> wrote:
> Some first impressions in glancing at the code, not a complete review at
> all:
>
> On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
>> --- /dev/null
>> +++ b/drivers/coresight/coresight.c
>> @@ -0,0 +1,663 @@
>> +/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
>> + *
>> + * This program is free software; you can redistribute it and/or modify
>> + * it under the terms of the GNU General Public License version 2 and
>> + * only version 2 as published by the Free Software Foundation.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> + * GNU General Public License for more details.
>> + */
>> +
>> +#define pr_fmt(fmt) "coresight: " fmt
>
> MODULE_NAME ?

Is this what you had in mind?

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

>
>> +
>> +#include <linux/kernel.h>
>> +#include <linux/module.h>
>> +#include <linux/init.h>
>> +#include <linux/types.h>
>> +#include <linux/device.h>
>> +#include <linux/io.h>
>> +#include <linux/err.h>
>> +#include <linux/export.h>
>> +#include <linux/slab.h>
>> +#include <linux/semaphore.h>
>> +#include <linux/clk.h>
>> +#include <linux/coresight.h>
>> +#include <linux/of_platform.h>
>> +#include <linux/debugfs.h>
>> +#include <linux/delay.h>
>> +
>> +#include "coresight-priv.h"
>> +
>> +struct dentry *cs_debugfs_parent = NULL;
>> +
>> +static LIST_HEAD(coresight_orph_conns);
>> +static LIST_HEAD(coresight_devs);
>
> You are a struct device, you don't need a separate list, why not just
> iterate over your bus list of devices?

Because coresight devices are not powered on the bus.  As such, every
time we iterate over the bus we'd need to setup the clock for the
device, power up its domain (if necessary) and lookup the device ID.
Isn't better to simply keep a list of the devices that were found at
boot time and use that whenever we want to make configuration changes?

With regards to @coresight_orph_conns, on top of the above problem
we'd also have to add a flag in the @coresight_device structure to
indicate that a device has been connected to the topology or not. To
me using a list is much cleaner.

>
>> +/**
>> + * @id:              unique ID of the component.
>> + * @conns:   list of connections associated to this component.
>> + * @type:    as defined by @coresight_dev_type.
>> + * @subtype: as defined by @coresight_dev_subtype.
>> + * @ops:     generic operations for this component, as defined
>> +             by @coresight_ops.
>> + * @de:              handle on a component's debugfs entry.
>> + * @dev:     The device entity associated to this component.
>> + * @kref:    keeping count on component references.
>> + * @dev_link:        link of current component into list of all components
>> +             discovered in the system.
>> + * @path_link:       link of current component into the path being enabled.
>> + * @owner:   typically "THIS_MODULE".
>> + * @enable:  'true' if component is currently part of an active path.
>> + * @activated:       'true' only if a _sink_ has been activated.  A sink can be
>> +             activated but not yet enabled.  Enabling for a _sink_
>> +             happens when a source has been selected for that it.
>> + */
>> +struct coresight_device {
>> +     int id;
>
> Why not use the device name instead?

With regards to memory footprint it is better to keep a single "int"
as ID (which is the their memory map start address) than the full
string associated with the device name.  Let's take a funnel for
example that has up to 8 input port.  To build a path between source
and sink we need to keep information about device that are connected
to each port.  At this time we use the component's ID, i.e 4 byte.  If
we were to use device names, ex "20040000.funnel_cluster0", we are
using 24 byte and that, 8 times.

Moreover using device names would also mean that we have to used
@stcmp everytime we want do lookups.  To me using a single integer to
identify coresight components is a much better solution.

>
>> +     struct coresight_connection *conns;
>> +     int nr_conns;
>> +     enum coresight_dev_type type;
>> +     struct coresight_dev_subtype subtype;
>> +     const struct coresight_ops *ops;
>> +     struct dentry *de;
>
> All devices have a dentry?  in debugfs?  isn't that overkill?

All devices along a "path" can require specific configuration, which
can only be made by a user.  Exposing the registers via debugfs seemed
like the most plausible solution.

>
>> +     struct device dev;
>> +     struct kref kref;
>
> You CAN NOT have two reference counts in the same structure, that's a
> huge design mistake.  Stick with one reference count, otherwise they are
> guaranteed to get out of sync and bad things will happen.

In this case the additional @kref is for accounting of components
within the coresight framework only.  When the amba framework calls
the driver's _probe() routine the kref count is already equal to '2'
(as expected), even if no other coresight components have used that
device.  Knowing when a component is no longer in use by the framework
(but still available in the system) is important for coresight cleanup
consideration, i.e switch off the component to save power.

The kref helpers don't expose the refcount and @kref_sub will only
call the cleanup method when refcount is '1'.  How else should I
proceed?

>
>> +     struct list_head dev_link;
>
> As discussed above, I don't think you need this.
>
>> +     struct list_head path_link;
>
> Odds are, you don't need this either.
>
>> +     struct module *owner;
>
> devices aren't owned by modules, they are data, not code.

Ok

>
>
>> +     bool enable;    /* true only if configured as part of a path */
>> +     bool activated; /* true only if a sink is part of a path */
>> +};
>> +
>> +#define to_coresight_device(d) container_of(d, struct coresight_device, dev)
>> +
>> +#define source_ops(csdev)    csdev->ops->source_ops
>> +#define sink_ops(csdev)              csdev->ops->sink_ops
>> +#define link_ops(csdev)              csdev->ops->link_ops
>> +
>> +#define CORESIGHT_DEBUGFS_ENTRY(__name, __entry_name,                        \
>> +                              __mode, __get, __set, __fmt)           \
>> +DEFINE_SIMPLE_ATTRIBUTE(__name ## _ops, __get, __set, __fmt)         \
>
> Use the RW and RO only versions please.  No need to ever set your own
> mode value.

I think I get your point but not quite sure how you want me to go
about it.  Can you give me a couple more line pls?

Again, very grateful for the review.

>
> thanks,
>
> greg k-h
Greg Kroah-Hartman Sept. 12, 2014, 6:16 p.m. UTC | #3
On Fri, Sep 12, 2014 at 11:41:44AM -0600, Mathieu Poirier wrote:
> Good morning and thanks for the review.  Pls see comments below.
> 
> Mathieu
> 
> On 11 September 2014 14:33, Greg KH <gregkh@linuxfoundation.org> wrote:
> > Some first impressions in glancing at the code, not a complete review at
> > all:
> >
> > On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
> >> --- /dev/null
> >> +++ b/drivers/coresight/coresight.c
> >> @@ -0,0 +1,663 @@
> >> +/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
> >> + *
> >> + * This program is free software; you can redistribute it and/or modify
> >> + * it under the terms of the GNU General Public License version 2 and
> >> + * only version 2 as published by the Free Software Foundation.
> >> + *
> >> + * This program is distributed in the hope that it will be useful,
> >> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> >> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> >> + * GNU General Public License for more details.
> >> + */
> >> +
> >> +#define pr_fmt(fmt) "coresight: " fmt
> >
> > MODULE_NAME ?
> 
> Is this what you had in mind?
> 
> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

Sorry, yes, that's much better and "portable".

But as you are using the driver model, the odds of you ever using a pr_*
call is all but none, so why do you even need this?

> 
> >
> >> +
> >> +#include <linux/kernel.h>
> >> +#include <linux/module.h>
> >> +#include <linux/init.h>
> >> +#include <linux/types.h>
> >> +#include <linux/device.h>
> >> +#include <linux/io.h>
> >> +#include <linux/err.h>
> >> +#include <linux/export.h>
> >> +#include <linux/slab.h>
> >> +#include <linux/semaphore.h>
> >> +#include <linux/clk.h>
> >> +#include <linux/coresight.h>
> >> +#include <linux/of_platform.h>
> >> +#include <linux/debugfs.h>
> >> +#include <linux/delay.h>
> >> +
> >> +#include "coresight-priv.h"
> >> +
> >> +struct dentry *cs_debugfs_parent = NULL;
> >> +
> >> +static LIST_HEAD(coresight_orph_conns);
> >> +static LIST_HEAD(coresight_devs);
> >
> > You are a struct device, you don't need a separate list, why not just
> > iterate over your bus list of devices?
> 
> Because coresight devices are not powered on the bus.  As such, every
> time we iterate over the bus we'd need to setup the clock for the
> device, power up its domain (if necessary) and lookup the device ID.
> Isn't better to simply keep a list of the devices that were found at
> boot time and use that whenever we want to make configuration changes?

You have that list already, as part of the bus list in the driver core.
I'm not asking you to "power up" anything, just use the data structures
the kernel already provides for you.

> With regards to @coresight_orph_conns, on top of the above problem
> we'd also have to add a flag in the @coresight_device structure to
> indicate that a device has been connected to the topology or not. To
> me using a list is much cleaner.

A single flag is "cleaner" than a whole separate list, not to mention
simpler to understand :)

> >> +/**
> >> + * @id:              unique ID of the component.
> >> + * @conns:   list of connections associated to this component.
> >> + * @type:    as defined by @coresight_dev_type.
> >> + * @subtype: as defined by @coresight_dev_subtype.
> >> + * @ops:     generic operations for this component, as defined
> >> +             by @coresight_ops.
> >> + * @de:              handle on a component's debugfs entry.
> >> + * @dev:     The device entity associated to this component.
> >> + * @kref:    keeping count on component references.
> >> + * @dev_link:        link of current component into list of all components
> >> +             discovered in the system.
> >> + * @path_link:       link of current component into the path being enabled.
> >> + * @owner:   typically "THIS_MODULE".
> >> + * @enable:  'true' if component is currently part of an active path.
> >> + * @activated:       'true' only if a _sink_ has been activated.  A sink can be
> >> +             activated but not yet enabled.  Enabling for a _sink_
> >> +             happens when a source has been selected for that it.
> >> + */
> >> +struct coresight_device {
> >> +     int id;
> >
> > Why not use the device name instead?
> 
> With regards to memory footprint it is better to keep a single "int"
> as ID (which is the their memory map start address) than the full
> string associated with the device name.  Let's take a funnel for
> example that has up to 8 input port.  To build a path between source
> and sink we need to keep information about device that are connected
> to each port.  At this time we use the component's ID, i.e 4 byte.  If
> we were to use device names, ex "20040000.funnel_cluster0", we are
> using 24 byte and that, 8 times.
> 
> Moreover using device names would also mean that we have to used
> @stcmp everytime we want do lookups.  To me using a single integer to
> identify coresight components is a much better solution.

Don't try to optimize for something that isn't an issue at all.  You
aren't searching for devices based on id in any time-critical section,
and you already have the device id string, so you aren't saving any
memory, but the opposite, wasting some.  See, your "optimization" ended
up taking more memory :)

> >> +     struct coresight_connection *conns;
> >> +     int nr_conns;
> >> +     enum coresight_dev_type type;
> >> +     struct coresight_dev_subtype subtype;
> >> +     const struct coresight_ops *ops;
> >> +     struct dentry *de;
> >
> > All devices have a dentry?  in debugfs?  isn't that overkill?
> 
> All devices along a "path" can require specific configuration, which
> can only be made by a user.  Exposing the registers via debugfs seemed
> like the most plausible solution.

configuration is not to be done through debugfs, especially as that is
restricted to root, and lots of systems don't even have it.  debugfs is
always optional, don't make your code depend on it to work properly.

> >> +     struct device dev;
> >> +     struct kref kref;
> >
> > You CAN NOT have two reference counts in the same structure, that's a
> > huge design mistake.  Stick with one reference count, otherwise they are
> > guaranteed to get out of sync and bad things will happen.
> 
> In this case the additional @kref is for accounting of components
> within the coresight framework only.  When the amba framework calls
> the driver's _probe() routine the kref count is already equal to '2'
> (as expected), even if no other coresight components have used that
> device.  Knowing when a component is no longer in use by the framework
> (but still available in the system) is important for coresight cleanup
> consideration, i.e switch off the component to save power.
> 
> The kref helpers don't expose the refcount and @kref_sub will only
> call the cleanup method when refcount is '1'.  How else should I
> proceed?

Don't use a kref at all, you should never need to "know" a refcount
value.  If you do, your logic is incorrect, sorry.  Just use the normal
driver core functions for when to cleanup memory and the like and you
will be fine.

> >> +#define to_coresight_device(d) container_of(d, struct coresight_device, dev)
> >> +
> >> +#define source_ops(csdev)    csdev->ops->source_ops
> >> +#define sink_ops(csdev)              csdev->ops->sink_ops
> >> +#define link_ops(csdev)              csdev->ops->link_ops
> >> +
> >> +#define CORESIGHT_DEBUGFS_ENTRY(__name, __entry_name,                        \
> >> +                              __mode, __get, __set, __fmt)           \
> >> +DEFINE_SIMPLE_ATTRIBUTE(__name ## _ops, __get, __set, __fmt)         \
> >
> > Use the RW and RO only versions please.  No need to ever set your own
> > mode value.
> 
> I think I get your point but not quite sure how you want me to go
> about it.  Can you give me a couple more line pls?

Why not just use DEVICE_ATTR_RW()?  and DEVICE_ATTR_RO()? Why create
your own attribute type?  And if you do, at the very least, use
__ATTR_RO() and __ATTR_RW(), don't roll your own macros.

And again, you should not need any other mode values other than RO or
RW, don't make a developer have to think about mode values for any
attributes.

thanks,

greg k-h
Russell King - ARM Linux Sept. 12, 2014, 6:44 p.m. UTC | #4
Further to Greg's comments...

On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
> +int coresight_enable(struct coresight_device *csdev)
> +{
> +	int ret = 0;
> +	LIST_HEAD(path);
> +
> +	WARN_ON(IS_ERR_OR_NULL(csdev));

Please don't do this kind of checking, it just makes stuff a lot more
noisy than it needs to be, and it doesn't give any value what so ever.

I've seen code where it seems the coding style required that quite
literally every function does extensive checking of function arguments,
and every function returns a status.  This does nothing to stop bugs.
In fact, it makes things a /lot/ worse because there is then soo much
junk in every function that you can't read what the function is doing
anymore.  Imagine memset() validating its arguments and returning a
status value... I kid not.

The point here is that if a NULL pointer is passed to this function,
the above WARN_ON gets triggered, and we get a backtrace.  We then
continue on, take the semaphore, and then dereference the NULL pointer
causing another backtrace.  So the WARN_ON was utterly pointless.

Just reference the pointer and don't bother with these silly checks.

> +
> +	down(&coresight_semaphore);

What is your reason for using a semaphore rather than a mutex?

...
> +/**
> + * coresight_timeout - loop until a bit has changed to a specific state.
> + * @addr: base address of the area of interest.
> + * @offset: address of a register, starting from @addr.
> + * @position: the position of the bit of interest.
> + * @value: the value the bit should have.
> + *
> + * Returns as soon as the bit has taken the desired state or TIMEOU_US has

Typo?

> + * elapsed, which ever happens first.
> + */
> +
> +void coresight_timeout(void __iomem *addr, u32 offset, int position, int value)
> +{
> +	int i;
> +	u32 val;
> +
> +	for (i = TIMEOUT_US; i > 0; i--) {
> +		val = __raw_readl(addr + offset);
> +		/* waiting on the bit to go from 0 to 1 */
> +		if (value) {
> +			if (val & BIT(position))
> +				return;
> +		/* waiting on the bit to go from 1 to 0 */
> +		} else {
> +			if (!(val & BIT(position)))
> +				return;
> +		}
> +
> +		/* The specification doesn't say how long we are expected
> +		 * to wait.
> +		 */

		/*
		 * The kernel commeting style for multi-line comments is
		 * like this.  Note the line opening the comment has no
		 * comment text.
		 */

> +		udelay(1);

So the duration is arbitary.

> +	}
> +
> +	WARN(1,
> +	     "coresight: timeout observed when proving at offset %#x\n",
> +	     offset);

This is also buggy.  On the last loop iteration, we delay 1us, decrement
i, and then test for it being greater than zero.  If it isn't, print
this message and do a backtrace (why is a backtrace useful here?)
The important point here is that we waited for 1us, and /didn't/ test
for success before claiming timeout.  That makes the final 1us wait
entirely useless.

> diff --git a/drivers/coresight/of_coresight.c b/drivers/coresight/of_coresight.c
> new file mode 100644
> index 0000000..c780b4b
> --- /dev/null
> +++ b/drivers/coresight/of_coresight.c
> @@ -0,0 +1,201 @@
...
> +struct coresight_platform_data *of_get_coresight_platform_data(
> +				struct device *dev, struct device_node *node)
> +{
> +	int id, i = 0, ret = 0;
> +	struct device_node *cpu;
> +	struct coresight_platform_data *pdata;
> +	struct of_endpoint endpoint, rendpoint;
> +	struct device_node *ep = NULL;
> +	struct device_node *rparent = NULL;
> +	struct device_node *rport = NULL;
> +
> +	pdata = devm_kzalloc(dev, sizeof(*pdata), GFP_KERNEL);
> +	if (!pdata)
> +		return ERR_PTR(-ENOMEM);

So, what are the rules for calling this function?  What is the expected
lifetime of this pdata structure in relation to 'dev' ?

You do realise that when a driver unbinds from 'dev', this allocation
will be freed.  If you hold on to the pointer and dereference it, you
could be accessing memory allocated for other purposes at that point.

> diff --git a/include/linux/coresight.h b/include/linux/coresight.h
> new file mode 100644
> index 0000000..5b22287
> --- /dev/null
> +++ b/include/linux/coresight.h
> @@ -0,0 +1,275 @@
...
> +/**
> + * @name:	name of the entry to appear under the component's
> +		debugfs sub-directory.
> + * @mode:	what operation can be performed on the entry.
> + * @ops:	specific manipulation to be done using this entry.
> + */

Wrong commenting style.  Please read Documentation/kernel-doc-nano-HOWTO.txt
for information how to document structures.
Mathieu Poirier Sept. 16, 2014, 7:42 p.m. UTC | #5
On 12 September 2014 12:44, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> Further to Greg's comments...
>
> On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
>> +int coresight_enable(struct coresight_device *csdev)
>> +{
>> +     int ret = 0;
>> +     LIST_HEAD(path);
>> +
>> +     WARN_ON(IS_ERR_OR_NULL(csdev));
>
> Please don't do this kind of checking, it just makes stuff a lot more
> noisy than it needs to be, and it doesn't give any value what so ever.
>
> I've seen code where it seems the coding style required that quite
> literally every function does extensive checking of function arguments,
> and every function returns a status.  This does nothing to stop bugs.
> In fact, it makes things a /lot/ worse because there is then soo much
> junk in every function that you can't read what the function is doing
> anymore.  Imagine memset() validating its arguments and returning a
> status value... I kid not.
>
> The point here is that if a NULL pointer is passed to this function,
> the above WARN_ON gets triggered, and we get a backtrace.  We then
> continue on, take the semaphore, and then dereference the NULL pointer
> causing another backtrace.  So the WARN_ON was utterly pointless.
>
> Just reference the pointer and don't bother with these silly checks.

Very well.

>
>> +
>> +     down(&coresight_semaphore);
>
> What is your reason for using a semaphore rather than a mutex?

Thanks for insisting on this - after doing a little more research on
the topic I will be moving to a mutex implementation.

>
> ...
>> +/**
>> + * coresight_timeout - loop until a bit has changed to a specific state.
>> + * @addr: base address of the area of interest.
>> + * @offset: address of a register, starting from @addr.
>> + * @position: the position of the bit of interest.
>> + * @value: the value the bit should have.
>> + *
>> + * Returns as soon as the bit has taken the desired state or TIMEOU_US has
>
> Typo?

Indeed.

>
>> + * elapsed, which ever happens first.
>> + */
>> +
>> +void coresight_timeout(void __iomem *addr, u32 offset, int position, int value)
>> +{
>> +     int i;
>> +     u32 val;
>> +
>> +     for (i = TIMEOUT_US; i > 0; i--) {
>> +             val = __raw_readl(addr + offset);
>> +             /* waiting on the bit to go from 0 to 1 */
>> +             if (value) {
>> +                     if (val & BIT(position))
>> +                             return;
>> +             /* waiting on the bit to go from 1 to 0 */
>> +             } else {
>> +                     if (!(val & BIT(position)))
>> +                             return;
>> +             }
>> +
>> +             /* The specification doesn't say how long we are expected
>> +              * to wait.
>> +              */
>
>                 /*
>                  * The kernel commeting style for multi-line comments is
>                  * like this.  Note the line opening the comment has no
>                  * comment text.
>                  */
>
>> +             udelay(1);
>
> So the duration is arbitary.

Correct.

>
>> +     }
>> +
>> +     WARN(1,
>> +          "coresight: timeout observed when proving at offset %#x\n",
>> +          offset);
>
> This is also buggy.  On the last loop iteration, we delay 1us, decrement
> i, and then test for it being greater than zero.  If it isn't, print
> this message and do a backtrace (why is a backtrace useful here?)
> The important point here is that we waited for 1us, and /didn't/ test
> for success before claiming timeout.  That makes the final 1us wait
> entirely useless.

Good catch.

>
>> diff --git a/drivers/coresight/of_coresight.c b/drivers/coresight/of_coresight.c
>> new file mode 100644
>> index 0000000..c780b4b
>> --- /dev/null
>> +++ b/drivers/coresight/of_coresight.c
>> @@ -0,0 +1,201 @@
> ...
>> +struct coresight_platform_data *of_get_coresight_platform_data(
>> +                             struct device *dev, struct device_node *node)
>> +{
>> +     int id, i = 0, ret = 0;
>> +     struct device_node *cpu;
>> +     struct coresight_platform_data *pdata;
>> +     struct of_endpoint endpoint, rendpoint;
>> +     struct device_node *ep = NULL;
>> +     struct device_node *rparent = NULL;
>> +     struct device_node *rport = NULL;
>> +
>> +     pdata = devm_kzalloc(dev, sizeof(*pdata), GFP_KERNEL);
>> +     if (!pdata)
>> +             return ERR_PTR(-ENOMEM);
>
> So, what are the rules for calling this function?  What is the expected
> lifetime of this pdata structure in relation to 'dev' ?
>
> You do realise that when a driver unbinds from 'dev', this allocation
> will be freed.  If you hold on to the pointer and dereference it, you
> could be accessing memory allocated for other purposes at that point.

By the time the allocation of pdata is freed the driver will have
called "coresight_unregister()", removing all knowledge of that entity
by the framework.

>
>> diff --git a/include/linux/coresight.h b/include/linux/coresight.h
>> new file mode 100644
>> index 0000000..5b22287
>> --- /dev/null
>> +++ b/include/linux/coresight.h
>> @@ -0,0 +1,275 @@
> ...
>> +/**
>> + * @name:    name of the entry to appear under the component's
>> +             debugfs sub-directory.
>> + * @mode:    what operation can be performed on the entry.
>> + * @ops:     specific manipulation to be done using this entry.
>> + */
>
> Wrong commenting style.  Please read Documentation/kernel-doc-nano-HOWTO.txt
> for information how to document structures.

Ok.

>
> --
> FTTC broadband for 0.8mile line: currently at 9.5Mbps down 400kbps up
> according to speedtest.net.
Mathieu Poirier Sept. 18, 2014, 11:09 p.m. UTC | #6
On 12 September 2014 12:16, Greg KH <gregkh@linuxfoundation.org> wrote:
> On Fri, Sep 12, 2014 at 11:41:44AM -0600, Mathieu Poirier wrote:
>> Good morning and thanks for the review.  Pls see comments below.
>>
>> Mathieu
>>
>> On 11 September 2014 14:33, Greg KH <gregkh@linuxfoundation.org> wrote:
>> > Some first impressions in glancing at the code, not a complete review at
>> > all:
>> >
>> > On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
>> >> --- /dev/null
>> >> +++ b/drivers/coresight/coresight.c
>> >> @@ -0,0 +1,663 @@
>> >> +/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
>> >> + *
>> >> + * This program is free software; you can redistribute it and/or modify
>> >> + * it under the terms of the GNU General Public License version 2 and
>> >> + * only version 2 as published by the Free Software Foundation.
>> >> + *
>> >> + * This program is distributed in the hope that it will be useful,
>> >> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> >> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> >> + * GNU General Public License for more details.
>> >> + */
>> >> +
>> >> +#define pr_fmt(fmt) "coresight: " fmt
>> >
>> > MODULE_NAME ?
>>
>> Is this what you had in mind?
>>
>> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>
> Sorry, yes, that's much better and "portable".
>
> But as you are using the driver model, the odds of you ever using a pr_*
> call is all but none, so why do you even need this?

You suggest to use dev_* instead?  My (perhaps flawed) logic was that
the framework ins't associated with a single device but a set of them.

>> >
>> >> +
>> >> +#include <linux/kernel.h>
>> >> +#include <linux/module.h>
>> >> +#include <linux/init.h>
>> >> +#include <linux/types.h>
>> >> +#include <linux/device.h>
>> >> +#include <linux/io.h>
>> >> +#include <linux/err.h>
>> >> +#include <linux/export.h>
>> >> +#include <linux/slab.h>
>> >> +#include <linux/semaphore.h>
>> >> +#include <linux/clk.h>
>> >> +#include <linux/coresight.h>  The original author had created a new bus (called "coresight") and lumped all the coresight components found on the system under that sysfs entry.
>> >> +#include <linux/of_platform.h>
>> >> +#include <linux/debugfs.h>
>> >> +#include <linux/delay.h>
>> >> +
>> >> +#include "coresight-priv.h"
>> >> +
>> >> +struct dentry *cs_debugfs_parent = NULL;
>> >> +
>> >> +static LIST_HEAD(coresight_orph_conns);
>> >> +static LIST_HEAD(coresight_devs);
>> >
>> > You are a struct device, you don't need a separate list, why not just
>> > iterate over your bus list of devices?
>>
>> Because coresight devices are not powered on the bus.  As such, every
>> time we iterate over the bus we'd need to setup the clock for the
>> device, power up its domain (if necessary) and lookup the device ID.
>> Isn't better to simply keep a list of the devices that were found at
>> boot time and use that whenever we want to make configuration changes?
>
> You have that list already, as part of the bus list in the driver core.
> I'm not asking you to "power up" anything, just use the data structures
> the kernel already provides for you.

After looking into this further I believe to understand what you are asking for.

>
>> With regards to @coresight_orph_conns, on top of the above problem
>> we'd also have to add a flag in the @coresight_device structure to
>> indicate that a device has been connected to the topology or not. To
>> me using a list is much cleaner.
>
> A single flag is "cleaner" than a whole separate list, not to mention
> simpler to understand :)  The original author had created a new bus (called "coresight") and lumped all the coresight components found on the system under that sysfs entry.

Very well.

>
>> >> +/**
>> >> + * @id:              unique ID of the component.
>> >> + * @conns:   list of connections associated to this component.
>> >> + * @type:    as defined by @coresight_dev_type.
>> >> + * @subtype: as defined by @coresight_dev_subtype.
>> >> + * @ops:     generic operations for this component, as defined
>> >> +             by @coresight_ops.
>> >> + * @de:              handle on a component's debugfs entry.
>> >> + * @dev:     The device entity associated to this component.
>> >> + * @kref:    keeping count on component references.
>> >> + * @dev_link:        link of current component into list of all components
>> >> +             discovered in the system.
>> >> + * @path_link:       link of current component into the path being enabled.
>> >> + * @owner:   typically "THIS_MODULE".
>> >> + * @enable:  'true' if component is currently part of an active path.
>> >> + * @activated:       'true' only if a _sink_ has been activated.  A sink can be
>> >> +             activated but not yet enabled.  Enabling for a _sink_
>> >> +             happens when a source has been selected for that it.
>> >> + */
>> >> +struct coresight_device {
>> >> +     int id;
>> >
>> > Why not use the device name instead?
>>
>> With regards to memory footprint it is better to keep a single "int"
>> as ID (which is the their memory map start address) than the full
>> string associated with the device name.  Let's take a funnel for
>> example that has up to 8 input port.  To build a path between source
>> and sink we need to keep information about device that are connected
>> to each port.  At this time we use the component's ID, i.e 4 byte.  If
>> we were to use device names, ex "20040000.funnel_cluster0", we are
>> using 24 byte and that, 8 times.
>>
>> Moreover using device names would also mean that we have to used
>> @stcmp everytime we want do lookups.  To me using a single integer to
>> identify coresight components is a much better solution.
>
> Don't try to optimize for something that isn't an issue at all.  You
> aren't searching for devices based on id in any time-critical section,
> and you already have the device id string, so you aren't saving any
> memory, but the opposite, wasting some.  See, your "optimization" ended
> up taking more memory :)

Ok, the least I can do is to give it a try.

>
>> >> +     struct coresight_connection *conns;
>> >> +     int nr_conns;
>> >> +     enum coresight_dev_type type;
>> >> +     struct coresight_dev_subtype subtype;
>> >> +     const struct coresight_ops *ops;
>> >> +     struct dentry *de;
>> >
>> > All devices have a dentry?  in debugfs?  isn't that overkill?
>>
>> All devices along a "path" can require specific configuration, which
>> can only be made by a user.  Exposing the registers via debugfs seemed
>> like the most plausible solution.
>
> configuration is not to be done through debugfs, especially as that is
> restricted to root, and lots of systems don't even have it.  debugfs is
> always optional, don't make your code depend on it to work properly.

Humm... You are the first one to be of the opinion that debugfs
shouldn't be used for coresight configuration - other reviewers have
unequivocally requested that configurables be moved under debugfs.
Debugfs seemed to be a good fit, specifically since ftrace is already
in there.  We could enable CONFIG_DEBUG_FS automatically (the same way
I do it for AMBA) when coresight gets enabled in the kernel config...

If not debugfs, where then?  I'd welcome a suggestion.

>
>> >> +     struct device dev;
>> >> +     struct kref kref;
>> >
>> > You CAN NOT have two reference counts in the same structure, that's a
>> > huge design mistake.  Stick with one reference count, otherwise they are
>> > guaranteed to get out of sync and bad things will happen.
>>
>> In this case the additional @kref is for accounting of components
>> within the coresight framework only.  When the amba framework calls
>> the driver's _probe() routine the kref count is already equal to '2'
>> (as expected), even if no other coresight components have used that
>> device.  Knowing when a component is no longer in use by the framework
>> (but still available in the system) is important for coresight cleanup
>> consideration, i.e switch off the component to save power.
>>
>> The kref helpers don't expose the refcount and @kref_sub will only
>> call the cleanup method when refcount is '1'.  How else should I
>> proceed?
>
> Don't use a kref at all, you should never need to "know" a refcount
> value.  If you do, your logic is incorrect, sorry.  Just use the normal
> driver core functions for when to cleanup memory and the like and you
> will be fine.

I completely agree with you on the fact that the driver core alone
should be the one doing the memory cleanup for a device.  What those
krefs are doing is to keep track of what is happening in the
framework, nothing else.

The framework needs to keep track of component usage by other
components... Say we have a sink that is used by multiple sources.
The framework can't call sink->disable() for as long as a coresight
source is using it and this is exactly what those krefs are for.
Disabling a device in the framework doesn't mean the kernel should be
reclaiming resources for the device.  That device should still be
there for future processing if need be.  On the flip side if a
component isn't used by the framework there is no point using
resources for it i.e, clock, pinout lines, channels...

>
>> >> +#define to_coresight_device(d) container_of(d, struct coresight_device, dev)
>> >> +
>> >> +#define source_ops(csdev)    csdev->ops->source_ops
>> >> +#define sink_ops(csdev)              csdev->ops->sink_ops
>> >> +#define link_ops(csdev)              csdev->ops->link_ops
>> >> +
>> >> +#define CORESIGHT_DEBUGFS_ENTRY(__name, __entry_name,                        \
>> >> +                              __mode, __get, __set, __fmt)           \
>> >> +DEFINE_SIMPLE_ATTRIBUTE(__name ## _ops, __get, __set, __fmt)         \
>> >
>> > Use the RW and RO only versions please.  No need to ever set your own
>> > mode value.
>>
>> I think I get your point but not quite sure how you want me to go
>> about it.  Can you give me a couple more line pls?
>
> Why not just use DEVICE_ATTR_RW()?  and DEVICE_ATTR_RO()? Why create
> your own attribute type?  And if you do, at the very least, use
> __ATTR_RO() and __ATTR_RW(), don't roll your own macros.
>
> And again, you should not need any other mode values other than RO or
> RW, don't make a developer have to think about mode values for any
> attributes.
>
> thanks,
>
> greg k-h
Greg Kroah-Hartman Sept. 24, 2014, 5:45 a.m. UTC | #7
On Thu, Sep 18, 2014 at 05:09:11PM -0600, Mathieu Poirier wrote:
> On 12 September 2014 12:16, Greg KH <gregkh@linuxfoundation.org> wrote:
> > On Fri, Sep 12, 2014 at 11:41:44AM -0600, Mathieu Poirier wrote:
> >> Good morning and thanks for the review.  Pls see comments below.
> >>
> >> Mathieu
> >>
> >> On 11 September 2014 14:33, Greg KH <gregkh@linuxfoundation.org> wrote:
> >> > Some first impressions in glancing at the code, not a complete review at
> >> > all:
> >> >
> >> > On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
> >> >> --- /dev/null
> >> >> +++ b/drivers/coresight/coresight.c
> >> >> @@ -0,0 +1,663 @@
> >> >> +/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
> >> >> + *
> >> >> + * This program is free software; you can redistribute it and/or modify
> >> >> + * it under the terms of the GNU General Public License version 2 and
> >> >> + * only version 2 as published by the Free Software Foundation.
> >> >> + *
> >> >> + * This program is distributed in the hope that it will be useful,
> >> >> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> >> >> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> >> >> + * GNU General Public License for more details.
> >> >> + */
> >> >> +
> >> >> +#define pr_fmt(fmt) "coresight: " fmt
> >> >
> >> > MODULE_NAME ?
> >>
> >> Is this what you had in mind?
> >>
> >> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> >
> > Sorry, yes, that's much better and "portable".
> >
> > But as you are using the driver model, the odds of you ever using a pr_*
> > call is all but none, so why do you even need this?
> 
> You suggest to use dev_* instead?  My (perhaps flawed) logic was that
> the framework ins't associated with a single device but a set of them.

Yes, but a single device is causing a message to be emitted, so
reference that please.

> >> >> +     struct coresight_connection *conns;
> >> >> +     int nr_conns;
> >> >> +     enum coresight_dev_type type;
> >> >> +     struct coresight_dev_subtype subtype;
> >> >> +     const struct coresight_ops *ops;
> >> >> +     struct dentry *de;
> >> >
> >> > All devices have a dentry?  in debugfs?  isn't that overkill?
> >>
> >> All devices along a "path" can require specific configuration, which
> >> can only be made by a user.  Exposing the registers via debugfs seemed
> >> like the most plausible solution.
> >
> > configuration is not to be done through debugfs, especially as that is
> > restricted to root, and lots of systems don't even have it.  debugfs is
> > always optional, don't make your code depend on it to work properly.
> 
> Humm... You are the first one to be of the opinion that debugfs
> shouldn't be used for coresight configuration - other reviewers have
> unequivocally requested that configurables be moved under debugfs.
> Debugfs seemed to be a good fit, specifically since ftrace is already
> in there.  We could enable CONFIG_DEBUG_FS automatically (the same way
> I do it for AMBA) when coresight gets enabled in the kernel config...
> 
> If not debugfs, where then?  I'd welcome a suggestion.

It is a mistake that ftrace is in debugfs, the developers of that code
admit it.

Your system should be able to run just fine if debugfs is disabled, if
your interface is "optional" like that, great, put it in debugfs,
otherwise put it somewhere else.

As to where else to put it, what exactly are you trying to do here?
What do you want to export / import?  Who would use it?  How would they
use it?  When would they use it?

> >> >> +     struct device dev;
> >> >> +     struct kref kref;
> >> >
> >> > You CAN NOT have two reference counts in the same structure, that's a
> >> > huge design mistake.  Stick with one reference count, otherwise they are
> >> > guaranteed to get out of sync and bad things will happen.
> >>
> >> In this case the additional @kref is for accounting of components
> >> within the coresight framework only.  When the amba framework calls
> >> the driver's _probe() routine the kref count is already equal to '2'
> >> (as expected), even if no other coresight components have used that
> >> device.  Knowing when a component is no longer in use by the framework
> >> (but still available in the system) is important for coresight cleanup
> >> consideration, i.e switch off the component to save power.
> >>
> >> The kref helpers don't expose the refcount and @kref_sub will only
> >> call the cleanup method when refcount is '1'.  How else should I
> >> proceed?
> >
> > Don't use a kref at all, you should never need to "know" a refcount
> > value.  If you do, your logic is incorrect, sorry.  Just use the normal
> > driver core functions for when to cleanup memory and the like and you
> > will be fine.
> 
> I completely agree with you on the fact that the driver core alone
> should be the one doing the memory cleanup for a device.  What those
> krefs are doing is to keep track of what is happening in the
> framework, nothing else.

That's not what a kref is for, don't abuse it that way.  You shouldn't
care what is "happening in the framework" if the driver core is handling
the reference counting properly, right?

> The framework needs to keep track of component usage by other
> components...

No it should not.

> Say we have a sink that is used by multiple sources.
> The framework can't call sink->disable() for as long as a coresight
> source is using it and this is exactly what those krefs are for.
> Disabling a device in the framework doesn't mean the kernel should be
> reclaiming resources for the device.  That device should still be
> there for future processing if need be.

What do you mean "if need be"?  How could the resource come back?  Why
do you need to care if a device is "disabled" or not?

> On the flip side if a component isn't used by the framework there is
> no point using resources for it i.e, clock, pinout lines, channels...

Make up your mind :)

greg k-h
Mathieu Poirier Sept. 26, 2014, 9:23 p.m. UTC | #8
On 23 September 2014 23:45, Greg KH <gregkh@linuxfoundation.org> wrote:
> On Thu, Sep 18, 2014 at 05:09:11PM -0600, Mathieu Poirier wrote:
>> On 12 September 2014 12:16, Greg KH <gregkh@linuxfoundation.org> wrote:
>> > On Fri, Sep 12, 2014 at 11:41:44AM -0600, Mathieu Poirier wrote:
>> >> Good morning and thanks for the review.  Pls see comments below.
>> >>
>> >> Mathieu
>> >>
>> >> On 11 September 2014 14:33, Greg KH <gregkh@linuxfoundation.org> wrote:
>> >> > Some first impressions in glancing at the code, not a complete review at
>> >> > all:
>> >> >
>> >> > On Thu, Sep 11, 2014 at 09:49:08AM -0600, mathieu.poirier@linaro.org wrote:
>> >> >> --- /dev/null
>> >> >> +++ b/drivers/coresight/coresight.c
>> >> >> @@ -0,0 +1,663 @@
>> >> >> +/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
>> >> >> + *
>> >> >> + * This program is free software; you can redistribute it and/or modify
>> >> >> + * it under the terms of the GNU General Public License version 2 and
>> >> >> + * only version 2 as published by the Free Software Foundation.
>> >> >> + *
>> >> >> + * This program is distributed in the hope that it will be useful,
>> >> >> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> >> >> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> >> >> + * GNU General Public License for more details.
>> >> >> + */
>> >> >> +
>> >> >> +#define pr_fmt(fmt) "coresight: " fmt
>> >> >
>> >> > MODULE_NAME ?
>> >>
>> >> Is this what you had in mind?
>> >>
>> >> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>> >
>> > Sorry, yes, that's much better and "portable".
>> >
>> > But as you are using the driver model, the odds of you ever using a pr_*
>> > call is all but none, so why do you even need this?
>>
>> You suggest to use dev_* instead?  My (perhaps flawed) logic was that
>> the framework ins't associated with a single device but a set of them.
>
> Yes, but a single device is causing a message to be emitted, so
> reference that please.
>
>> >> >> +     struct coresight_connection *conns;
>> >> >> +     int nr_conns;
>> >> >> +     enum coresight_dev_type type;
>> >> >> +     struct coresight_dev_subtype subtype;
>> >> >> +     const struct coresight_ops *ops;
>> >> >> +     struct dentry *de;
>> >> >
>> >> > All devices have a dentry?  in debugfs?  isn't that overkill?
>> >>
>> >> All devices along a "path" can require specific configuration, which
>> >> can only be made by a user.  Exposing the registers via debugfs seemed
>> >> like the most plausible solution.
>> >
>> > configuration is not to be done through debugfs, especially as that is
>> > restricted to root, and lots of systems don't even have it.  debugfs is
>> > always optional, don't make your code depend on it to work properly.
>>
>> Humm... You are the first one to be of the opinion that debugfs
>> shouldn't be used for coresight configuration - other reviewers have
>> unequivocally requested that configurables be moved under debugfs.
>> Debugfs seemed to be a good fit, specifically since ftrace is already
>> in there.  We could enable CONFIG_DEBUG_FS automatically (the same way
>> I do it for AMBA) when coresight gets enabled in the kernel config...
>>
>> If not debugfs, where then?  I'd welcome a suggestion.
>
> It is a mistake that ftrace is in debugfs, the developers of that code
> admit it.

That's interesting to know.

>
> Your system should be able to run just fine if debugfs is disabled, if
> your interface is "optional" like that, great, put it in debugfs,
> otherwise put it somewhere else.

I lumped the entries under /sys/bus/coresight/, the same way the
orgininal author had them.

>
> As to where else to put it, what exactly are you trying to do here?
> What do you want to export / import?  Who would use it?  How would they
> use it?  When would they use it?
>
>> >> >> +     struct device dev;
>> >> >> +     struct kref kref;
>> >> >
>> >> > You CAN NOT have two reference counts in the same structure, that's a
>> >> > huge design mistake.  Stick with one reference count, otherwise they are
>> >> > guaranteed to get out of sync and bad things will happen.
>> >>
>> >> In this case the additional @kref is for accounting of components
>> >> within the coresight framework only.  When the amba framework calls
>> >> the driver's _probe() routine the kref count is already equal to '2'
>> >> (as expected), even if no other coresight components have used that
>> >> device.  Knowing when a component is no longer in use by the framework
>> >> (but still available in the system) is important for coresight cleanup
>> >> consideration, i.e switch off the component to save power.
>> >>
>> >> The kref helpers don't expose the refcount and @kref_sub will only
>> >> call the cleanup method when refcount is '1'.  How else should I
>> >> proceed?
>> >
>> > Don't use a kref at all, you should never need to "know" a refcount
>> > value.  If you do, your logic is incorrect, sorry.  Just use the normal
>> > driver core functions for when to cleanup memory and the like and you
>> > will be fine.
>>
>> I completely agree with you on the fact that the driver core alone
>> should be the one doing the memory cleanup for a device.  What those
>> krefs are doing is to keep track of what is happening in the
>> framework, nothing else.
>
> That's not what a kref is for, don't abuse it that way.

I was using ref_counts before but changing to krefs under the advice
of another maintainer.

> You shouldn't
> care what is "happening in the framework" if the driver core is handling
> the reference counting properly, right?

By framework I meant the coresight framework.  We need to keep track
of what is happening and what coresight components are active.  That
is all those krefs were doing.  Again, I'll gladly move back to using
ref_counts if I've abused the usage of krefs...

>
>> The framework needs to keep track of component usage by other
>> components...
>
> No it should not.
>
>> Say we have a sink that is used by multiple sources.
>> The framework can't call sink->disable() for as long as a coresight
>> source is using it and this is exactly what those krefs are for.
>> Disabling a device in the framework doesn't mean the kernel should be
>> reclaiming resources for the device.  That device should still be
>> there for future processing if need be.
>
> What do you mean "if need be"?  How could the resource come back?  Why
> do you need to care if a device is "disabled" or not?
>
>> On the flip side if a component isn't used by the framework there is
>> no point using resources for it i.e, clock, pinout lines, channels...
>
> Make up your mind :)
>
> greg k-h
diff mbox

Patch

diff --git a/arch/arm/Kconfig.debug b/arch/arm/Kconfig.debug
index 8f90595..c04d0d9 100644
--- a/arch/arm/Kconfig.debug
+++ b/arch/arm/Kconfig.debug
@@ -1257,4 +1257,13 @@  config DEBUG_SET_MODULE_RONX
 	  against certain classes of kernel exploits.
 	  If in doubt, say "N".
 
+menuconfig CORESIGHT
+	bool "CoreSight Tracing Support"
+	select ARM_AMBA
+	help
+	  This framework provides a kernel interface for the CoreSight debug
+	  and trace drivers to register themselves with. It's intended to build
+	  a topological view of the CoreSight components based on a DT
+	  specification and configure the right serie of components when a
+	  trace source gets enabled.
 endmenu
diff --git a/drivers/Makefile b/drivers/Makefile
index f98b50d..fb58a94 100644
--- a/drivers/Makefile
+++ b/drivers/Makefile
@@ -158,3 +158,4 @@  obj-$(CONFIG_NTB)		+= ntb/
 obj-$(CONFIG_FMC)		+= fmc/
 obj-$(CONFIG_POWERCAP)		+= powercap/
 obj-$(CONFIG_MCB)		+= mcb/
+obj-$(CONFIG_CORESIGHT)		+= coresight/
diff --git a/drivers/amba/bus.c b/drivers/amba/bus.c
index 3cf61a1..131258b 100644
--- a/drivers/amba/bus.c
+++ b/drivers/amba/bus.c
@@ -327,7 +327,7 @@  int amba_device_add(struct amba_device *dev, struct resource *parent)
 
 		amba_put_disable_pclk(dev);
 
-		if (cid == AMBA_CID)
+		if (cid == AMBA_CID || cid == CORESIGHT_CID)
 			dev->periphid = pid;
 
 		if (!dev->periphid)
diff --git a/drivers/coresight/Makefile b/drivers/coresight/Makefile
new file mode 100644
index 0000000..218e3b5
--- /dev/null
+++ b/drivers/coresight/Makefile
@@ -0,0 +1,5 @@ 
+#
+# Makefile for CoreSight drivers.
+#
+obj-$(CONFIG_CORESIGHT) += coresight.o
+obj-$(CONFIG_OF) += of_coresight.o
diff --git a/drivers/coresight/coresight-priv.h b/drivers/coresight/coresight-priv.h
new file mode 100644
index 0000000..83172c4
--- /dev/null
+++ b/drivers/coresight/coresight-priv.h
@@ -0,0 +1,63 @@ 
+/* Copyright (c) 2011-2012, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _CORESIGHT_PRIV_H
+#define _CORESIGHT_PRIV_H
+
+#include <linux/bitops.h>
+#include <linux/io.h>
+#include <linux/coresight.h>
+
+/*
+ * Coresight management registers (0xf00-0xfcc)
+ * 0xfa0 - 0xfa4: Management	registers in PFTv1.0
+ *		  Trace		registers in PFTv1.1
+ */
+#define CORESIGHT_ITCTRL	0xf00
+#define CORESIGHT_CLAIMSET	0xfa0
+#define CORESIGHT_CLAIMCLR	0xfa4
+#define CORESIGHT_LAR		0xfb0
+#define CORESIGHT_LSR		0xfb4
+#define CORESIGHT_AUTHSTATUS	0xfb8
+#define CORESIGHT_DEVID		0xfc8
+#define CORESIGHT_DEVTYPE	0xfcc
+
+#define TIMEOUT_US		100
+#define BMVAL(val, lsb, msb)	((val & GENMASK(msb, lsb)) >> lsb)
+
+static inline void CS_LOCK(void __iomem *addr)
+{
+	do {
+		/* wait for things to settle */
+		mb();
+		writel_relaxed(0x0, addr + CORESIGHT_LAR);
+	} while (0);
+}
+
+static inline void CS_UNLOCK(void __iomem *addr)
+{
+	do {
+		writel_relaxed(CORESIGHT_UNLOCK, addr + CORESIGHT_LAR);
+		/* make sure eveyone has seen this */
+		mb();
+	} while (0);
+}
+
+#ifdef CONFIG_CORESIGHT_SOURCE_ETM3X
+extern unsigned int etm_readl_cp14(u32 off);
+extern void etm_writel_cp14(u32 val, u32 off);
+#else
+static inline unsigned int etm_readl_cp14(u32 off) { return 0; }
+static inline void etm_writel_cp14(u32 val, u32 off) {}
+#endif
+
+#endif
diff --git a/drivers/coresight/coresight.c b/drivers/coresight/coresight.c
new file mode 100644
index 0000000..7e1e6b7
--- /dev/null
+++ b/drivers/coresight/coresight.c
@@ -0,0 +1,663 @@ 
+/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#define pr_fmt(fmt) "coresight: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/device.h>
+#include <linux/io.h>
+#include <linux/err.h>
+#include <linux/export.h>
+#include <linux/slab.h>
+#include <linux/semaphore.h>
+#include <linux/clk.h>
+#include <linux/coresight.h>
+#include <linux/of_platform.h>
+#include <linux/debugfs.h>
+#include <linux/delay.h>
+
+#include "coresight-priv.h"
+
+struct dentry *cs_debugfs_parent = NULL;
+
+static LIST_HEAD(coresight_orph_conns);
+static LIST_HEAD(coresight_devs);
+static DEFINE_SEMAPHORE(coresight_semaphore);
+
+static int coresight_source_is_unique(struct coresight_device *csdev)
+{
+	struct coresight_device *cd;
+	int trace_id = source_ops(csdev)->trace_id(csdev);
+
+	/* this shouldn't happen */
+	if (trace_id < 0)
+		return 0;
+
+	/* circle through all known components looking for sources */
+	list_for_each_entry(cd, &coresight_devs, dev_link) {
+		/* no need to care about oneself and components that are not
+		 * sources or not enabled
+		 */
+		if (cd == csdev || !cd->enable ||
+		    cd->type != CORESIGHT_DEV_TYPE_SOURCE)
+			continue;
+
+		/* all you need is one */
+		if (trace_id == source_ops(cd)->trace_id(cd))
+			return 0;
+	}
+
+	return 1;
+}
+
+static int coresight_find_link_inport(struct coresight_device *csdev)
+{
+	int i;
+	struct coresight_device *parent;
+	struct coresight_connection *conn;
+
+	parent = container_of(csdev->path_link.next, struct coresight_device,
+			     path_link);
+	for (i = 0; i < parent->nr_conns; i++) {
+		conn = &parent->conns[i];
+		if (conn->child_dev == csdev)
+			return conn->child_port;
+	}
+
+	pr_err("couldn't find inport, parent: %d, child: %d\n",
+	       parent->id, csdev->id);
+	return 0;
+}
+
+static int coresight_find_link_outport(struct coresight_device *csdev)
+{
+	int i;
+	struct coresight_device *child;
+	struct coresight_connection *conn;
+
+	child = container_of(csdev->path_link.prev, struct coresight_device,
+			      path_link);
+	for (i = 0; i < csdev->nr_conns; i++) {
+		conn = &csdev->conns[i];
+		if (conn->child_dev == child)
+			return conn->outport;
+	}
+
+	pr_err("couldn't find outport, parent: %d, child: %d\n",
+	       csdev->id, child->id);
+	return 0;
+}
+
+static int coresight_enable_sink(struct coresight_device *csdev)
+{
+	int ret;
+
+	if (!csdev->enable) {
+		if (sink_ops(csdev)->enable) {
+			ret = sink_ops(csdev)->enable(csdev);
+			if (ret)
+				return ret;
+		}
+		kref_init(&csdev->kref);
+		csdev->enable = true;
+		return 0;
+	}
+
+	kref_get(&csdev->kref);
+
+	return 0;
+}
+
+static void coresight_driver_disable_sink(struct kref *kref)
+{
+	struct coresight_device *csdev = container_of(kref,
+					struct coresight_device, kref);
+
+	if (sink_ops(csdev)->disable) {
+		sink_ops(csdev)->disable(csdev);
+		csdev->enable = false;
+	}
+}
+
+static void coresight_disable_sink(struct coresight_device *csdev)
+{
+	kref_put(&csdev->kref, coresight_driver_disable_sink);
+}
+
+static int coresight_enable_link(struct coresight_device *csdev)
+{
+	int ret;
+	int inport, outport;
+
+	inport = coresight_find_link_inport(csdev);
+	outport = coresight_find_link_outport(csdev);
+
+	if (link_ops(csdev)->enable) {
+		ret = link_ops(csdev)->enable(csdev, inport, outport);
+		if (ret)
+			return ret;
+	}
+
+	if (!csdev->enable) {
+		csdev->enable = true;
+		kref_init(&csdev->kref);
+	} else {
+		kref_get(&csdev->kref);
+	}
+
+	return 0;
+}
+
+static void coresight_driver_disable_link(struct kref *kref)
+{
+	struct coresight_device *csdev = container_of(kref,
+					struct coresight_device, kref);
+	csdev->enable = false;
+}
+
+static void coresight_disable_link(struct coresight_device *csdev)
+{
+	int inport, outport;
+
+	inport = coresight_find_link_inport(csdev);
+	outport = coresight_find_link_outport(csdev);
+
+	if (link_ops(csdev)->disable)
+		link_ops(csdev)->disable(csdev, inport, outport);
+
+	kref_put(&csdev->kref, coresight_driver_disable_link);
+}
+
+static int coresight_enable_source(struct coresight_device *csdev)
+{
+	int ret;
+
+	if (!coresight_source_is_unique(csdev)) {
+		pr_warn("traceID %d not unique\n",
+			source_ops(csdev)->trace_id(csdev));
+		return -EINVAL;
+	}
+
+	if (!csdev->enable) {
+		if (source_ops(csdev)->enable) {
+			ret = source_ops(csdev)->enable(csdev);
+			if (ret)
+				return ret;
+		}
+		kref_init(&csdev->kref);
+		csdev->enable = true;
+		return 0;
+	}
+
+	kref_get(&csdev->kref);
+
+	return 0;
+}
+
+static void coresight_driver_disable_source(struct kref *kref)
+{
+	struct coresight_device *csdev = container_of(kref,
+					struct coresight_device, kref);
+
+	if (source_ops(csdev)->disable) {
+		source_ops(csdev)->disable(csdev);
+		csdev->enable = false;
+	}
+}
+
+static void coresight_disable_source(struct coresight_device *csdev)
+{
+	kref_put(&csdev->kref, coresight_driver_disable_source);
+}
+
+static int coresight_enable_path(struct list_head *path)
+{
+	int ret = 0;
+	struct coresight_device *cd;
+
+	list_for_each_entry(cd, path, path_link) {
+		if (cd == list_first_entry(path, struct coresight_device,
+					   path_link)) {
+			ret = coresight_enable_sink(cd);
+		} else if (list_is_last(&cd->path_link, path)) {
+			/* dont' enable the source just yet - this needs to
+			 * happen at the very end when all links and sink
+			 * along the path have been configured properly.
+			 */
+			;
+		} else {
+			ret = coresight_enable_link(cd);
+		}
+		if (ret)
+			goto err;
+	}
+	return 0;
+err:
+	list_for_each_entry_continue_reverse(cd, path, path_link) {
+		if (cd == list_first_entry(path, struct coresight_device,
+					   path_link)) {
+			coresight_disable_sink(cd);
+		} else if (list_is_last(&cd->path_link, path)) {
+			;
+		} else {
+			coresight_disable_link(cd);
+		}
+	}
+	return ret;
+}
+
+static int coresight_disable_path(struct list_head *path)
+{
+	struct coresight_device *cd;
+
+	list_for_each_entry_reverse(cd, path, path_link) {
+		if (cd == list_first_entry(path, struct coresight_device,
+					   path_link)) {
+			coresight_disable_sink(cd);
+		} else if (list_is_last(&cd->path_link, path)) {
+			/* the source has already been stopped, no need
+			 * to do it again here.
+			 */
+			;
+		} else {
+			coresight_disable_link(cd);
+		}
+	}
+
+	return 0;
+}
+
+static int coresight_build_paths(struct coresight_device *csdev,
+				 struct list_head *path,
+				 bool enable)
+{
+	int i, ret = -EINVAL;
+	struct coresight_connection *conn;
+
+	list_add(&csdev->path_link, path);
+
+	if (csdev->type == CORESIGHT_DEV_TYPE_SINK && csdev->activated) {
+		if (enable)
+			ret = coresight_enable_path(path);
+		else
+			ret = coresight_disable_path(path);
+	} else {
+		for (i = 0; i < csdev->nr_conns; i++) {
+			conn = &csdev->conns[i];
+			if (coresight_build_paths(conn->child_dev,
+						    path, enable) == 0)
+				ret = 0;
+		}
+	}
+
+	if (list_first_entry(path, struct coresight_device, path_link) != csdev)
+		pr_err("wrong device in %s\n", __func__);
+
+	list_del(&csdev->path_link);
+	return ret;
+}
+
+int coresight_enable(struct coresight_device *csdev)
+{
+	int ret = 0;
+	LIST_HEAD(path);
+
+	WARN_ON(IS_ERR_OR_NULL(csdev));
+
+	down(&coresight_semaphore);
+	if (csdev->type != CORESIGHT_DEV_TYPE_SOURCE) {
+		ret = -EINVAL;
+		pr_err("wrong device type in %s\n", __func__);
+		goto out;
+	}
+	if (csdev->enable)
+		goto out;
+
+	if (coresight_build_paths(csdev, &path, true)) {
+		pr_err("building path(s) failed\n");
+		goto out;
+	}
+
+	if (coresight_enable_source(csdev))
+		pr_err("source enable failed\n");
+out:
+	up(&coresight_semaphore);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(coresight_enable);
+
+void coresight_disable(struct coresight_device *csdev)
+{
+	LIST_HEAD(path);
+
+	WARN_ON(IS_ERR_OR_NULL(csdev));
+
+	down(&coresight_semaphore);
+	if (csdev->type != CORESIGHT_DEV_TYPE_SOURCE) {
+		pr_err("wrong device type in %s\n", __func__);
+		goto out;
+	}
+	if (!csdev->enable)
+		goto out;
+
+	coresight_disable_source(csdev);
+	if (coresight_build_paths(csdev, &path, false))
+		pr_err("releasing path(s) failed\n");
+
+out:
+	up(&coresight_semaphore);
+}
+EXPORT_SYMBOL_GPL(coresight_disable);
+
+static ssize_t debugfs_active_get(void *data, u64 *val)
+{
+	struct coresight_device *csdev = data;
+
+	*val = csdev->activated;
+	return 0;
+}
+
+static ssize_t debugfs_active_set(void *data, u64 val)
+{
+	struct coresight_device *csdev = data;
+
+	csdev->activated = !!val;
+	return 0;
+}
+CORESIGHT_DEBUGFS_ENTRY(debugfs_active, "enable",
+			S_IRUGO | S_IWUSR, debugfs_active_get,
+			debugfs_active_set, "%llx\n");
+
+static ssize_t debugfs_enable_get(void *data, u64 *val)
+{
+	struct coresight_device *csdev = data;
+
+	*val = csdev->enable;
+	return 0;
+}
+
+static ssize_t debugfs_enable_set(void *data, u64 val)
+{
+	struct coresight_device *csdev = data;
+
+	if (val)
+		return coresight_enable(csdev);
+	else
+		coresight_disable(csdev);
+
+	return 0;
+}
+CORESIGHT_DEBUGFS_ENTRY(debugfs_enable, "enable",
+			S_IRUGO | S_IWUSR, debugfs_enable_get,
+			debugfs_enable_set, "%llx\n");
+
+
+static const struct coresight_ops_entry *coresight_grps_sink[] = {
+	&debugfs_active_entry,
+	NULL,
+};
+
+static const struct coresight_ops_entry *coresight_grps_source[] = {
+	&debugfs_enable_entry,
+	NULL,
+};
+
+struct coresight_group_entries {
+	const char *name;
+	const struct coresight_ops_entry **entries;
+};
+
+struct coresight_group_entries coresight_debugfs_entries[] = {
+	{
+		.name = "none",
+	},
+	{
+		.name = "sink",
+		.entries = coresight_grps_sink,
+	},
+	{
+		.name = "link",
+	},
+	{
+		.name = "linksink",
+	},
+	{
+		.name = "source",
+		.entries = coresight_grps_source,
+	},
+};
+
+static void coresight_device_release(struct device *dev)
+{
+	struct coresight_device *csdev = to_coresight_device(dev);
+
+	kfree(csdev);
+}
+
+static void coresight_fixup_orphan_conns(struct coresight_device *csdev)
+{
+	struct coresight_connection *conn, *temp;
+
+	list_for_each_entry_safe(conn, temp, &coresight_orph_conns, link) {
+		if (conn->child_id == csdev->id) {
+			conn->child_dev = csdev;
+			list_del(&conn->link);
+		}
+	}
+}
+
+static void coresight_fixup_device_conns(struct coresight_device *csdev)
+{
+	int i;
+	struct coresight_device *cd;
+	bool found;
+
+	for (i = 0; i < csdev->nr_conns; i++) {
+		found = false;
+		list_for_each_entry(cd, &coresight_devs, dev_link) {
+			if (csdev->conns[i].child_id == cd->id) {
+				csdev->conns[i].child_dev = cd;
+				found = true;
+				break;
+			}
+		}
+		if (!found)
+			list_add_tail(&csdev->conns[i].link,
+				      &coresight_orph_conns);
+	}
+}
+
+static int debugfs_coresight_init(void)
+{
+	if (!cs_debugfs_parent) {
+		cs_debugfs_parent = debugfs_create_dir("coresight", 0);
+		if (IS_ERR(cs_debugfs_parent))
+			return PTR_ERR(cs_debugfs_parent);
+	}
+
+	return 0;
+}
+
+static struct dentry *coresight_debugfs_desc_init(
+				struct coresight_device *csdev,
+				const struct coresight_ops_entry **debugfs_ops)
+{
+	int i = 0;
+	struct dentry *parent;
+	struct device *dev = &csdev->dev;
+	const struct coresight_ops_entry *ops_entry, **ops_entries;
+
+	parent = debugfs_create_dir(dev_name(dev), cs_debugfs_parent);
+	if (IS_ERR(parent))
+		return NULL;
+
+	/* device-specific ops */
+	while (debugfs_ops && debugfs_ops[i]) {
+		ops_entry = debugfs_ops[i];
+		if (!debugfs_create_file(ops_entry->name, ops_entry->mode,
+					 parent, dev_get_drvdata(dev->parent),
+					 ops_entry->ops)) {
+			debugfs_remove_recursive(parent);
+			return NULL;
+		}
+		i++;
+	}
+
+	/* group-specific ops */
+	i = 0;
+	ops_entries = coresight_debugfs_entries[csdev->type].entries;
+
+	while (ops_entries && ops_entries[i]) {
+		if (!debugfs_create_file(ops_entries[i]->name,
+					 ops_entries[i]->mode,
+					 parent, csdev, ops_entries[i]->ops)) {
+			debugfs_remove_recursive(parent);
+			return NULL;
+		}
+		i++;
+	}
+
+	return parent;
+}
+
+/**
+ * coresight_timeout - loop until a bit has changed to a specific state.
+ * @addr: base address of the area of interest.
+ * @offset: address of a register, starting from @addr.
+ * @position: the position of the bit of interest.
+ * @value: the value the bit should have.
+ *
+ * Returns as soon as the bit has taken the desired state or TIMEOU_US has
+ * elapsed, which ever happens first.
+ */
+
+void coresight_timeout(void __iomem *addr, u32 offset, int position, int value)
+{
+	int i;
+	u32 val;
+
+	for (i = TIMEOUT_US; i > 0; i--) {
+		val = __raw_readl(addr + offset);
+		/* waiting on the bit to go from 0 to 1 */
+		if (value) {
+			if (val & BIT(position))
+				return;
+		/* waiting on the bit to go from 1 to 0 */
+		} else {
+			if (!(val & BIT(position)))
+				return;
+		}
+
+		/* The specification doesn't say how long we are expected
+		 * to wait.
+		 */
+		udelay(1);
+	}
+
+	WARN(1,
+	     "coresight: timeout observed when proving at offset %#x\n",
+	     offset);
+}
+
+struct coresight_device *coresight_register(struct coresight_desc *desc)
+{
+	int i;
+	int ret;
+	int link_subtype;
+	struct coresight_device *csdev;
+	struct coresight_connection *conns;
+
+	WARN_ON(IS_ERR_OR_NULL(desc));
+
+	csdev = kzalloc(sizeof(*csdev), GFP_KERNEL);
+	if (!csdev) {
+		ret = -ENOMEM;
+		goto err_kzalloc_csdev;
+	}
+
+	csdev->id = desc->pdata->id;
+
+	if (desc->type == CORESIGHT_DEV_TYPE_LINK ||
+	    desc->type == CORESIGHT_DEV_TYPE_LINKSINK)
+		link_subtype = desc->subtype.link_subtype;
+
+	csdev->nr_conns = desc->pdata->nr_outports;
+	conns = kcalloc(csdev->nr_conns, sizeof(*conns), GFP_KERNEL);
+	if (!conns) {
+		ret = -ENOMEM;
+		goto err_kzalloc_conns;
+	}
+
+	for (i = 0; i < csdev->nr_conns; i++) {
+		conns[i].outport = desc->pdata->outports[i];
+		conns[i].child_id = desc->pdata->child_ids[i];
+		conns[i].child_port = desc->pdata->child_ports[i];
+	}
+	csdev->conns = conns;
+
+	csdev->type = desc->type;
+	csdev->subtype = desc->subtype;
+	csdev->ops = desc->ops;
+	csdev->owner = desc->owner;
+
+	csdev->dev.parent = desc->dev;
+	csdev->dev.release = coresight_device_release;
+	dev_set_name(&csdev->dev, "%s", desc->pdata->name);
+
+
+	down(&coresight_semaphore);
+
+	coresight_fixup_device_conns(csdev);
+
+	ret = debugfs_coresight_init();
+	if (ret < 0)
+		goto err_coresight_init;
+
+	csdev->de = coresight_debugfs_desc_init(csdev, desc->debugfs_ops);
+
+	coresight_fixup_orphan_conns(csdev);
+
+	list_add_tail(&csdev->dev_link, &coresight_devs);
+	up(&coresight_semaphore);
+
+	return csdev;
+
+err_coresight_init:
+	up(&coresight_semaphore);
+	kfree(conns);
+err_kzalloc_conns:
+	kfree(csdev);
+err_kzalloc_csdev:
+	return ERR_PTR(ret);
+}
+EXPORT_SYMBOL_GPL(coresight_register);
+
+void coresight_unregister(struct coresight_device *csdev)
+{
+	WARN_ON(IS_ERR_OR_NULL(csdev));
+
+	down(&coresight_semaphore);
+
+	list_del(&csdev->dev_link);
+	debugfs_remove_recursive(csdev->de);
+	kfree(csdev->conns);
+	if (list_empty(&coresight_devs))
+		kfree(cs_debugfs_parent);
+
+	up(&coresight_semaphore);
+}
+EXPORT_SYMBOL_GPL(coresight_unregister);
+
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/coresight/of_coresight.c b/drivers/coresight/of_coresight.c
new file mode 100644
index 0000000..c780b4b
--- /dev/null
+++ b/drivers/coresight/of_coresight.c
@@ -0,0 +1,201 @@ 
+/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/clk.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_graph.h>
+#include <linux/coresight.h>
+#include <asm/smp_plat.h>
+
+static int of_get_coresight_id(struct device_node *node, int *id)
+{
+	const __be32 *reg;
+	u64 addr;
+
+	/* derive component id from its memory map */
+	reg = of_get_property(node, "reg", NULL);
+	if (reg) {
+		addr = of_translate_address(node, reg);
+		if (addr != OF_BAD_ADDR) {
+			*id = addr;
+			return 0;
+		}
+	}
+
+	/* no "reg", we have a non-configurable replicator */
+	reg = of_get_property(node, "id", NULL);
+	if (reg) {
+		*id = of_read_ulong(reg, 1);
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+static struct device_node *of_get_coresight_endpoint(
+		const struct device_node *parent, struct device_node *prev)
+{
+	struct device_node *node = of_graph_get_next_endpoint(parent, prev);
+
+	of_node_put(prev);
+	return node;
+}
+
+static void of_coresight_get_ports(struct device_node *node,
+				   int *nr_inports, int *nr_outports)
+{
+	struct device_node *ep = NULL;
+	int in = 0, out = 0;
+
+	do {
+		ep = of_get_coresight_endpoint(node, ep);
+		if (!ep)
+			break;
+
+		if (of_property_read_bool(ep, "slave-mode"))
+			in++;
+		else
+			out++;
+
+	} while (ep);
+
+	*nr_inports = in;
+	*nr_outports = out;
+}
+
+static int of_coresight_alloc_memory(struct device *dev,
+			struct coresight_platform_data *pdata)
+{
+	/* list of output port on this component */
+	pdata->outports = devm_kzalloc(dev, pdata->nr_outports *
+				       sizeof(*pdata->outports),
+				       GFP_KERNEL);
+	if (!pdata->outports)
+		return -ENOMEM;
+
+
+	/* children connected to this component via @outport */
+	pdata->child_ids = devm_kzalloc(dev, pdata->nr_outports *
+					sizeof(*pdata->child_ids),
+					GFP_KERNEL);
+	if (!pdata->child_ids)
+		return -ENOMEM;
+
+	/* port number on the child this component is connected to */
+	pdata->child_ports = devm_kzalloc(dev, pdata->nr_outports *
+					  sizeof(*pdata->child_ports),
+					  GFP_KERNEL);
+	if (!pdata->child_ports)
+		return -ENOMEM;
+
+	return 0;
+}
+
+struct coresight_platform_data *of_get_coresight_platform_data(
+				struct device *dev, struct device_node *node)
+{
+	int id, i = 0, ret = 0;
+	struct device_node *cpu;
+	struct coresight_platform_data *pdata;
+	struct of_endpoint endpoint, rendpoint;
+	struct device_node *ep = NULL;
+	struct device_node *rparent = NULL;
+	struct device_node *rport = NULL;
+
+	pdata = devm_kzalloc(dev, sizeof(*pdata), GFP_KERNEL);
+	if (!pdata)
+		return ERR_PTR(-ENOMEM);
+
+	/* use the base address as id */
+	ret = of_get_coresight_id(node, &id);
+	if (ret)
+		return ERR_PTR(ret);
+	pdata->id = id;
+
+	/* use device name as debugfs handle */
+	pdata->name = dev_name(dev);
+
+	/* get the number of input and output port for this component */
+	of_coresight_get_ports(node, &pdata->nr_inports, &pdata->nr_outports);
+
+	if (pdata->nr_outports) {
+		ret = of_coresight_alloc_memory(dev, pdata);
+		if (ret)
+			return ERR_PTR(ret);
+
+		/* iterate through each port to discover topology */
+		do {
+			/* get a handle on a port */
+			ep = of_get_coresight_endpoint(node, ep);
+			if (!ep)
+				break;
+
+			/* no need to deal with input ports, processing for as
+			 * processing for output ports will deal with them.
+			 */
+			if (of_find_property(ep, "slave-mode", NULL))
+				continue;
+
+			/* get a handle on the local endpoint */
+			ret = of_graph_parse_endpoint(ep, &endpoint);
+
+			if (ret)
+				continue;
+
+			/* the local out port number */
+			pdata->outports[i] = endpoint.id;
+
+			/* get a handle the remote port and parent
+			 * attached to it.
+			 */
+			rparent = of_graph_get_remote_port_parent(ep);
+			rport = of_graph_get_remote_port(ep);
+
+			if (!rparent || !rport)
+				continue;
+
+			if (of_graph_parse_endpoint(rport, &rendpoint))
+				continue;
+
+			ret = of_get_coresight_id(rparent, &id);
+			if (ret)
+				continue;
+			pdata->child_ids[i] = id;
+			pdata->child_ports[i] = rendpoint.id;
+
+			i++;
+		} while (ep);
+	}
+
+	/* affinity defaults to CPU0 */
+	pdata->cpu = 0;
+	cpu = of_parse_phandle(node, "cpu", 0);
+	if (cpu) {
+		const u32 *mpidr;
+		int len, index;
+
+		mpidr = of_get_property(cpu, "reg", &len);
+		if (mpidr && len == 4) {
+			index = get_logical_index(be32_to_cpup(mpidr));
+			if (index != -EINVAL)
+				pdata->cpu = index;
+		}
+	}
+
+	return pdata;
+}
+EXPORT_SYMBOL_GPL(of_get_coresight_platform_data);
diff --git a/include/linux/amba/bus.h b/include/linux/amba/bus.h
index fdd7e1b..cdddabe 100644
--- a/include/linux/amba/bus.h
+++ b/include/linux/amba/bus.h
@@ -23,6 +23,7 @@ 
 
 #define AMBA_NR_IRQS	9
 #define AMBA_CID	0xb105f00d
+#define CORESIGHT_CID	0xb105900d
 
 struct clk;
 
diff --git a/include/linux/coresight.h b/include/linux/coresight.h
new file mode 100644
index 0000000..5b22287
--- /dev/null
+++ b/include/linux/coresight.h
@@ -0,0 +1,275 @@ 
+/* Copyright (c) 2012, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _LINUX_CORESIGHT_H
+#define _LINUX_CORESIGHT_H
+
+#include <linux/device.h>
+
+/* Peripheral id registers (0xFD0-0xFEC) */
+#define CORESIGHT_PERIPHIDR4	0xfd0
+#define CORESIGHT_PERIPHIDR5	0xfd4
+#define CORESIGHT_PERIPHIDR6	0xfd8
+#define CORESIGHT_PERIPHIDR7	0xfdC
+#define CORESIGHT_PERIPHIDR0	0xfe0
+#define CORESIGHT_PERIPHIDR1	0xfe4
+#define CORESIGHT_PERIPHIDR2	0xfe8
+#define CORESIGHT_PERIPHIDR3	0xfeC
+/* Component id registers (0xFF0-0xFFC) */
+#define CORESIGHT_COMPIDR0	0xff0
+#define CORESIGHT_COMPIDR1	0xff4
+#define CORESIGHT_COMPIDR2	0xff8
+#define CORESIGHT_COMPIDR3	0xffC
+
+#define ETM_ARCH_V3_3		0x23
+#define ETM_ARCH_V3_5		0x25
+#define PFT_ARCH_V1_0		0x30
+#define PFT_ARCH_V1_1		0x31
+
+#define CORESIGHT_UNLOCK	0xc5acce55
+
+enum coresight_dev_type {
+	CORESIGHT_DEV_TYPE_NONE,
+	CORESIGHT_DEV_TYPE_SINK,
+	CORESIGHT_DEV_TYPE_LINK,
+	CORESIGHT_DEV_TYPE_LINKSINK,
+	CORESIGHT_DEV_TYPE_SOURCE,
+};
+
+enum coresight_dev_subtype_sink {
+	CORESIGHT_DEV_SUBTYPE_SINK_NONE,
+	CORESIGHT_DEV_SUBTYPE_SINK_PORT,
+	CORESIGHT_DEV_SUBTYPE_SINK_BUFFER,
+};
+
+enum coresight_dev_subtype_link {
+	CORESIGHT_DEV_SUBTYPE_LINK_NONE,
+	CORESIGHT_DEV_SUBTYPE_LINK_MERG,
+	CORESIGHT_DEV_SUBTYPE_LINK_SPLIT,
+	CORESIGHT_DEV_SUBTYPE_LINK_FIFO,
+};
+
+enum coresight_dev_subtype_source {
+	CORESIGHT_DEV_SUBTYPE_SOURCE_NONE,
+	CORESIGHT_DEV_SUBTYPE_SOURCE_PROC,
+	CORESIGHT_DEV_SUBTYPE_SOURCE_BUS,
+	CORESIGHT_DEV_SUBTYPE_SOURCE_SOFTWARE,
+};
+
+/**
+ * @name:	name of the entry to appear under the component's
+		debugfs sub-directory.
+ * @mode:	what operation can be performed on the entry.
+ * @ops:	specific manipulation to be done using this entry.
+ */
+struct coresight_ops_entry {
+	const char *name;
+	umode_t mode;
+	const struct file_operations *ops;
+};
+
+/**
+ * @sink_subtype:	type of sink this component is, as defined
+			by @coresight_dev_subtype_sink.
+ * @link_subtype:	type of link this component is, as defined
+			by @coresight_dev_subtype_link.
+ * @source_subtype:	type of source this component is, as defined
+			by @coresight_dev_subtype_source.
+ */
+struct coresight_dev_subtype {
+	enum coresight_dev_subtype_sink sink_subtype;
+	enum coresight_dev_subtype_link link_subtype;
+	enum coresight_dev_subtype_source source_subtype;
+};
+
+/**
+ * @id:		unique ID of the component, usually the memory mapped address.
+ * @cpu:	the CPU a source belongs to. Only applicable for ETM/PTMs.
+ * @name:	name of the component as shown under debugfs.
+ * @nr_inports:	number of input ports for this component.
+ * @nr_outports:number of output ports for this component.
+ * @child_ids:	unique ID of a child component.
+ * @child_ports:child component port number the current component is
+		connected  to.
+ * @clk:	The clock this component is associated to.
+ */
+struct coresight_platform_data {
+	int id;
+	int cpu;
+	const char *name;
+	int nr_inports;
+	int *outports;
+	int *child_ids;
+	int *child_ports;
+	int nr_outports;
+	struct clk *clk;
+};
+
+/**
+ * @type:	as defined by @coresight_dev_type.
+ * @subtype:	as defined by @coresight_dev_subtype.
+ * @ops:	generic operations for this component, as defined
+		by @coresight_ops.
+ * @pdata:	platform data collected from DT.
+ * @dev:	The device entity associated to this component.
+ * @debugfs_ops:operations specific to this component. These will end up
+		in the component's debugfs sub-directory.
+ * @owner:	typically "THIS_MODULE".
+ */
+struct coresight_desc {
+	enum coresight_dev_type type;
+	struct coresight_dev_subtype subtype;
+	const struct coresight_ops *ops;
+	struct coresight_platform_data *pdata;
+	struct device *dev;
+	const struct coresight_ops_entry **debugfs_ops;
+	struct module *owner;
+};
+
+/**
+ * @outport:	a connection's output port number.
+ * @chid_id:	remote component's ID.
+ * @child_port:	remote component's port number @output is connected to.
+ * @child_dev:	a @coresight_device representation of the component
+		connected to @outport.
+ * @link:	hook to @coresight_orph_conns, for framework internal
+		use only.
+ */
+struct coresight_connection {
+	int outport;
+	int child_id;
+	int child_port;
+	struct coresight_device *child_dev;
+	struct list_head link;
+};
+
+/**
+ * @id:		unique ID of the component.
+ * @conns:	list of connections associated to this component.
+ * @type:	as defined by @coresight_dev_type.
+ * @subtype:	as defined by @coresight_dev_subtype.
+ * @ops:	generic operations for this component, as defined
+		by @coresight_ops.
+ * @de:		handle on a component's debugfs entry.
+ * @dev:	The device entity associated to this component.
+ * @kref:	keeping count on component references.
+ * @dev_link:	link of current component into list of all components
+		discovered in the system.
+ * @path_link:	link of current component into the path being enabled.
+ * @owner:	typically "THIS_MODULE".
+ * @enable:	'true' if component is currently part of an active path.
+ * @activated:	'true' only if a _sink_ has been activated.  A sink can be
+		activated but not yet enabled.  Enabling for a _sink_
+		happens when a source has been selected for that it.
+ */
+struct coresight_device {
+	int id;
+	struct coresight_connection *conns;
+	int nr_conns;
+	enum coresight_dev_type type;
+	struct coresight_dev_subtype subtype;
+	const struct coresight_ops *ops;
+	struct dentry *de;
+	struct device dev;
+	struct kref kref;
+	struct list_head dev_link;
+	struct list_head path_link;
+	struct module *owner;
+	bool enable;	/* true only if configured as part of a path */
+	bool activated;	/* true only if a sink is part of a path */
+};
+
+#define to_coresight_device(d) container_of(d, struct coresight_device, dev)
+
+#define source_ops(csdev)	csdev->ops->source_ops
+#define sink_ops(csdev)		csdev->ops->sink_ops
+#define link_ops(csdev)		csdev->ops->link_ops
+
+#define CORESIGHT_DEBUGFS_ENTRY(__name, __entry_name,			\
+				 __mode, __get, __set, __fmt)		\
+DEFINE_SIMPLE_ATTRIBUTE(__name ## _ops, __get, __set, __fmt)		\
+static const struct coresight_ops_entry __name ## _entry = {		\
+	.name = __entry_name,						\
+	.mode = __mode,							\
+	.ops  = &__name ## _ops						\
+}
+
+/**
+ * Operations available for sinks
+ * @enable:	enables the sink.
+ * @disable:	disables the sink.
+ */
+struct coresight_ops_sink {
+	int (*enable)(struct coresight_device *csdev);
+	void (*disable)(struct coresight_device *csdev);
+};
+
+/**
+ * Operations available for links.
+ * @enable:	enables flow between iport and oport.
+ * @disable:	disables flow between iport and oport.
+ */
+struct coresight_ops_link {
+	int (*enable)(struct coresight_device *csdev, int iport, int oport);
+	void (*disable)(struct coresight_device *csdev, int iport, int oport);
+};
+
+/**
+ * Operations available for sources.
+ * @trace_id:	returns the value of the component's trace ID as known
+		to the HW.
+ * @enable:	enables tracing from a source.
+ * @disable:	disables tracing for a source.
+ */
+struct coresight_ops_source {
+	int (*trace_id)(struct coresight_device *csdev);
+	int (*enable)(struct coresight_device *csdev);
+	void (*disable)(struct coresight_device *csdev);
+};
+
+struct coresight_ops {
+	const struct coresight_ops_sink *sink_ops;
+	const struct coresight_ops_link *link_ops;
+	const struct coresight_ops_source *source_ops;
+};
+
+#ifdef CONFIG_CORESIGHT
+extern struct coresight_device *
+coresight_register(struct coresight_desc *desc);
+extern void coresight_unregister(struct coresight_device *csdev);
+extern int coresight_enable(struct coresight_device *csdev);
+extern void coresight_disable(struct coresight_device *csdev);
+extern int coresight_is_bit_set(u32 val, int position, int value);
+extern void coresight_timeout(void __iomem *addr, u32 offset,
+			      int position, int value);
+#ifdef CONFIG_OF
+extern struct coresight_platform_data *of_get_coresight_platform_data(
+				struct device *dev, struct device_node *node);
+#endif
+#else
+static inline struct coresight_device *
+coresight_register(struct coresight_desc *desc) { return NULL; }
+static inline void coresight_unregister(struct coresight_device *csdev) {}
+static inline int
+coresight_enable(struct coresight_device *csdev) { return -ENOSYS; }
+static inline void coresight_disable(struct coresight_device *csdev) {}
+static inline int coresight_is_bit_set(u32 val, int position, int value)
+					 { return 0; }
+static inline void coresight_timeout(void __iomem *addr, u32 offset,
+				     int position, int value) {}
+#ifdef CONFIG_OF
+static inline struct coresight_platform_data *of_get_coresight_platform_data(
+	struct device *dev, struct device_node *node) { return NULL; }
+#endif
+#endif
+
+#endif