if we allow compiler reorder our writes, we could
fall into a situation where dev->buf_len is reset
for no apparent reason.

This bug was found with a simple script which would
transfer data to an i2c client from 1 to 1024 bytes
(a simple for loop), when we got to transfer sizes
bigger than the fifo size, dev->buf_len was reset
to zero before we had an oportunity to handle XDR
Interrupt. Because dev->buf_len was zero, we entered
omap_i2c_transmit_data() to transfer zero bytes,
which would mean we would just silently exit
omap_i2c_transmit_data() without actually writing
anything to DATA register. That would cause XDR
IRQ to trigger forever and we would never transfer
the remaining bytes.

After adding the memory barrier, we also drop resetting
dev->buf_len to zero in omap_i2c_xfer_msg() because
both omap_i2c_transmit_data() and omap_i2c_receive_data()
will act until dev->buf_len reaches zero, rendering the
other write in omap_i2c_xfer_msg() redundant.

This patch has been tested with pandaboard for a few
iterations of the script mentioned above.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

This reverts commit 3db11fef
(ARM: OMAP: convert I2C driver to PM QoS for MPU latency constraints).
This commit causes I2C timeouts to appear on several OMAP3430/3530-based
boards:

  http://marc.info/?l=linux-arm-kernel&m=135071372426971&w=2
  http://marc.info/?l=linux-arm-kernel&m=135067558415214&w=2
  http://marc.info/?l=linux-arm-kernel&m=135216013608196&w=2

and appears to have been sent for merging before one of its prerequisites
was merged:

  http://marc.info/?l=linux-arm-kernel&m=135219411617621&w=2Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Acked-by: NJean Pihet <j-pihet@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Convert the driver from the outdated omap_pm_set_max_mpu_wakeup_lat
API to the new PM QoS API.
Since the constraint is on the MPU subsystem, use the PM_QOS_CPU_DMA_LATENCY
class of PM QoS. The resulting MPU constraints are used by cpuidle to
decide the next power state of the MPU subsystem.

The I2C device latency timing is derived from the FIFO size and the
clock speed and so is applicable to all OMAP SoCs.
Signed-off-by: NJean Pihet <j-pihet@ti.com>
Acked-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>
Acked-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

When booting using a device tree, the adapter number is dynamically
assigned after the log message is sent.
This patch modifies the log message to get a correct adapter id.
Signed-off-by: NFlorian Vaussard <florian.vaussard@epfl.ch>
Tested-by: NBenoit Cousson <b-cousson@ti.com>
Acked-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

move the goto out label one line down, so that
it can be used when stat is read as zero. All
other exits, can be done with a break statement.

While at that, also break out as soon as we
complete draining IRQ, since at that time
we know we transferred everything there was
to be transferred.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

this helps us reduce unnecessary pm transitions
in case we have another i2c message starting soon.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

before starting any messages we call pm_runtime_get_sync()
which will make sure that by the time we program a transfer
and our IRQ handler gets called, we're not suspended
anymore.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

for OMAP2, we can easily switch over to threaded
IRQs on the I2C driver. This will allow us to
spend less time in hardirq context.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
[Trivial formating changes]
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Currently omap_i2c_ack_stat doesn't use the stat variable.
After the read of the I2C_STAT_REG it is not used.
Remove the redundant read of the status register.
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

we can ack stat and complete the command from
the errata handling itself.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

This patch will try to avoid the usage of
draining feature by reconfiguring the FIFO
the start condition of each message based
on the message's size.

By doing that, we will be better utilizing
the FIFO when doing big transfers.

While at that also drop the now unneeded
check for dev->buf_len as we always know
the amount of data to be transmitted.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

instead of having multiple return points, use
a goto statement to make that clearer.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Always return IRQ_HANDLED otherwise we could get our IRQ line disabled due
to many spurious IRQs.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
[Trivial changes to commitlogs]
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

omap_i2c_dev is allocated with kzalloc(),
so we need not initialize b_hw to zero.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

that way we can ignore TX IRQs while in receiver
mode and ignore RX IRQs while in transmitter mode.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
[Remove unnecessary braces]
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

that's a nice helper from drivers core which
will give us the exact IRQ number, instead
of a pointer to an IRQ resource.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

According to flow diagrams on OMAP TRMs,
we should ACK the IRQ as they happen.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
[Ack the stat OMAP_I2C_STAT_AL in case of arbitration lost]
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

this will make sure that we execute at least once.
No functional changes otherwise.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

re-factor the common parts to a separate function,
so that code is easier to read and understand.

No functional changes.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Make it not depend on ISR's local variables
in order to make it easier to re-factor the
transmit data loop.
Also since we are waiting for XUDF(Transmitter underflow) just before
writing data lets not flag the underflow.
This is anyways going to go once we write
the data.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

While they do pretty much the same thing, there
are a few peculiarities. Specially WRT erratas,
it's best to split those out and re-factor the
read/write loop to another function which both
cases call.

This last part will be done on another patch.

While at that, also avoid an unncessary register
read since dev->fifo_len will always contain the
correct amount of data to be transferred.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

stat & BIT(1) is the same as BIT(1), so let's
simplify things a bit by removing "stat &" from
all omap_i2c_ack_stat() calls.

Code snippet (extremely simplified):

if (stat & NACK) {
        ...
        omap_i2c_ack_stat(dev, stat & NACK);
}

if (stat & RDR) {
        ...
        omap_i2c_ack_stat(dev, stat & RDR);
}

and so on. The tricky place is only WRT errata handling, for example:

if (*stat & (NACK | AL)) {
        omap_i2c_ack_stat(dev, *stat & (XRDY | XDR));
        ...
}

but in this case, the errata says we must clear XRDY and XDR if that
errata triggers, so if they just got enabled or not, it doesn't matter.

Another tricky place is RDR | RRDY (likewise for XDR | XRDY):

if (stat & (RDR | RRDY)) {
        ...
        omap_i2c_ack_stat(dev, stat & (RDR | RRDY));
}

again here there will be no issues because those IRQs never fire
simultaneously and one will only after after we have handled the
previous, that's because the same FIFO is used anyway and we won't shift
data into FIFO until we tell the IP "hey, I'm done with the FIFO, you
can shift more data"
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Reviewed-by : Santosh Shilimkar <santosh.shilimkar@ti.com>
[Added the explaination from the discurssion to the commit logs]
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

trivial patch to aid readability. No functional
changes.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

The patch intends to decrease the indentation level on the
data handling
by using the fact that else of if (dev->buf_len) is same as
if (!dev->buf_len)

if (dev->buf_len) {
	aaa;
} else {
	bbb;
	break;
}

to

if (!dev->buf_len) {
        bbb;
        break;
}
aaa;

Hence no functional changes.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Reviewed-by : Santosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

trivial patch, no functional changes

If the fifo is disabled or fifo_size is 0 the num_bytes
is set to 1. Else it is set to fifo_size or in case of a
draining interrupt the remaining bytes in the buff stat.
So the zero check is redundant and can be safely optimised.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Reviewed-by : Santosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

that helps deleting some boiler plate code
and lets driver-core manage our resources
for us.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

This prepares *of_device_id.data becoming const. Without this change
the following warning would occur:

	drivers/i2c/busses/i2c-omap.c: In function 'omap_i2c_probe':
	drivers/i2c/busses/i2c-omap.c:1025: warning: assignment discards qualifiers from pointer target type
Reviewed-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NUwe Kleine-König <u.kleine-koenig@pengutronix.de>

In omap_i2c_xfer(), ensure pm_runtime_put() is called, even on
failure.

Without this, after a failed xfer, the runtime PM usecount will have
been incremented, but not decremented causing the usecount to never
reach zero after a failure.  This keeps the device always runtime PM
enabled which keeps the enclosing power domain active, and prevents
full-chip retention/off from happening during idle.
Signed-off-by: NKevin Hilman <khilman@ti.com>
Reviewed-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Generate a stop condition after each message marked with I2C_M_STOP.

[JD: Add I2C_FUNC_PROTOCOL_MANGLING.]
Signed-off-by: NLaurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: NJean Delvare <khali@linux-fr.org>

The OMAP3530 is based upon the same silicon as the OMAP3430 and so the I2C
revision is the same for 3430 and 3530. However, the OMAP3630 device has the
same I2C revision as OMAP4. Correct the revision definition to reflect this.

This patch is based on work done by Jon Hunter <jon-hunter@ti.com>
Changes from his patch
- Update OMAP_I2C_REV_ON_3430 also to reflect that it is same as 3530
Reviewed-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NJon Hunter <jon-hunter@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Use INIT_COMPLETION instead of init_completion in transfer.
Reviewed-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Use SET_RUNTIME_PM_OPS macro to set runtime functions.
Acked-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

The omap_i2c_remove function may not be needed after
device exit so the memory could be freed.
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

On a board with OMAP3 processor and TWL4030 Power management,
we need to talk to the TWL4030 during late suspend but cannot
because the I2C interrupt is disabled (as late suspend disables
interrupt).

e.g. I get messages like:

[   62.161102] musb-omap2430 musb-omap2430: LATE power domain suspend
[   63.167205] omap_i2c omap_i2c.1: controller timed out
[   63.183044] twl: i2c_read failed to transfer all messages
[   64.182861] omap_i2c omap_i2c.1: controller timed out
[   64.198455] twl: i2c_write failed to transfer all messages
[   65.198455] omap_i2c omap_i2c.1: controller timed out
[   65.203765] twl: i2c_write failed to transfer all messages

The stack shows omap2430_runtime_suspend calling twl4030_set_suspend
which tries to power-down the USB PHY (twl4030_phy_suspend ->
twl4030_phy_power -> __twl4030_phy_power which as a nice WARN_ON
that helps).

Then we get the same in resume:

[   69.603912] musb-omap2430 musb-omap2430: EARLY power domain resume
[   70.610473] omap_i2c omap_i2c.1: controller timed out
[   70.626129] twl: i2c_write failed to transfer all messages
etc.

So don't disable interrupts for I2C.
Acked-by: NKevin Hilman <khilman@ti.com>
Tested-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NNeilBrown <neilb@suse.de>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

The section number in the recent errata document has changed.
Rename the erratum 1p153 to the unique id i462 instead, so that
it is easier to reference. Also change the function name and comments
to reflect the same.

Cc: Jon Hunter <jon-hunter@ti.com>
Reviewed-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

Currently in the 1.153 errata handling, while waiting for transmitter
underflow, if NACK is got the XUDF(Transmit underflow) flag is also set.
Fix this by setting the XUDF(Transmit underflow) flag after wait for the
condition is over.

Cc: Alexander Shishkin <virtuoso@slind.org>
Acked-by: NMoiz Sonasath <m-sonasath@ti.com>
Reviewed-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

i2c_probe set the dev->errata flag, but omap_i2c_init cleared the flag again.
Prevent the overwrite of the errata flags.Move the errata handling to a unified
place in probe to prevent such errors.
Reviewed-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NTasslehoff Kjappfot <tasskjapp@gmail.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

If PM runtime get_sync fails return with the error
so that no further reads/writes goes through the interface.
This will avoid possible abort. Add a error message in case
of failure with the cause of the failure.
Reviewed-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

    In omap_i2c_remove we are accessing the I2C_CON register without
    enabling the clocks. Fix the same by ensure device is accessible by calling
    pm_runtime_get_sync before accessing the registers and calling pm_runtime_put
    after accessing.

    This fixes the following crash.
    [  154.723022] ------------[ cut here ]------------
    [  154.725677] WARNING: at arch/arm/mach-omap2/omap_l3_noc.c:112 l3_interrupt_handler+0x1b4/0x1c4()
    [  154.725677] L3 custom error: MASTER:MPU TARGET:L4 PER2
    [  154.742614] Modules linked in: i2c_omap(-)
    [  154.746948] Backtrace:
    [  154.746948] [<c0013078>] (dump_backtrace+0x0/0x110) from [<c026c158>] (dump_stack+0x18/0x1c)
    [  154.752716]  r6:00000070 r5:c002c43c r4:df9b9e98 r3:df9b8000
    [  154.764465] [<c026c140>] (dump_stack+0x0/0x1c) from [<c0041a2c>] (warn_slowpath_common+0x5c/0x6c)
    [  154.768341] [<c00419d0>] (warn_slowpath_common+0x0/0x6c) from [<c0041ae0>] (warn_slowpath_fmt+0x38/0x40)
    [  154.776153]  r8:00000180 r7:c0361594 r6:c0379b48 r5:00080003 r4:e0838b00
    [  154.790771] r3:00000009
    [  154.791778] [<c0041aa8>] (warn_slowpath_fmt+0x0/0x40) from [<c002c43c>] (l3_interrupt_handler+0x1b4/0x1c4)
    [  154.803710]  r3:c0361598 r2:c02ef74c
    [  154.807403] [<c002c288>] (l3_interrupt_handler+0x0/0x1c4) from [<c0085f44>] (handle_irq_event_percpu+0x58/0
    [  154.818237]  r8:0000002a r7:00000000 r6:00000000 r5:df808054 r4:df8893c0
    [  154.825378] [<c0085eec>] (handle_irq_event_percpu+0x0/0x188) from [<c00860b8>] (handle_irq_event+0x44/0x64)
    [  154.835662] [<c0086074>] (handle_irq_event+0x0/0x64) from [<c0088ec0>] (handle_fasteoi_irq+0xa4/0x10c)
    [  154.845458]  r6:0000002a r5:df808054 r4:df808000 r3:c034a150
    [  154.846466] [<c0088e1c>] (handle_fasteoi_irq+0x0/0x10c) from [<c0085ed0>] (generic_handle_irq+0x30/0x38)
    [  154.854278]  r5:c034aa48 r4:0000002a
    [  154.862091] [<c0085ea0>] (generic_handle_irq+0x0/0x38) from [<c000fd38>] (handle_IRQ+0x60/0xc0)
    [  154.874450]  r4:c034ea70 r3:000001f8
    [  154.878234] [<c000fcd8>] (handle_IRQ+0x0/0xc0) from [<c0008478>] (gic_handle_irq+0x20/0x5c)
    [  154.887023]  r7:ffffff40 r6:df9b9fb0 r5:c034e2b4 r4:0000001a
    [  154.887054] [<c0008458>] (gic_handle_irq+0x0/0x5c) from [<c000ea80>] (__irq_usr+0x40/0x60)
    [  154.901153] Exception stack(0xdf9b9fb0 to 0xdf9b9ff8)
    [  154.907104] 9fa0:                                     beaf1f04 4006be00 0000000f 0000000c
    [  154.915710] 9fc0: 4006c000 00000000 00008034 ffffff40 00000007 00000000 00000000 0007b8d7
    [  154.916778] 9fe0: 00000000 beaf1b68 0000d23c 4005baf0 80000010 ffffffff
    [  154.931335]  r6:ffffffff r5:80000010 r4:4005baf0 r3:beaf1f04
    [  154.937316] ---[ end trace 1b75b31a2719ed21 ]--

Cc: Rajendra Nayak <rnayak@ti.com>
Cc: Linux PM list <linux-pm@vger.kernel.org>
Reviewed-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>

By definition, wait_for_completion_timeout() returns an unsigned value and
therefore, it is not necessary to check if the return value is less than zero
as this is not possible.

This is based on a patch from Jon Hunter <jon-hunter@ti.com>
Changes from his patch
- Declare a long as the wait_for_completion_timeout returns long.

Original patch is
http://git.omapzoom.org/?p=kernel/omap.git;a=commitdiff;h=ea02cece7b0000bc736e60c4188a11aaa74bc6e6Reviewed-by: NKevin Hilman <khilman@ti.com>
Signed-off-by: NJon Hunter <jon-hunter@ti.com>
Signed-off-by: NShubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: NWolfram Sang <w.sang@pengutronix.de>