The manual says that we need to (repeatedly) set the TearDown-bit for
the endpoint in order to get the active transfer descriptor released.
Doing this "real" quick over and over again seems to work but it also
seems that the hardware might not have enough time to breathe. So I
though, hey lets add a udelay() between between the individual sets
of the bit.
This change with the g_zero testcase resulted in a warning about missing
transfer descriptor (we got the tear-down one). It seems that if the
hardware has some time it manages to release the transfer-descriptor on
the completion queue after the teaddown descriptor.
With this change, I observe that the transfer descriptor is released
after 20-30 retry loops.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
For a "complete" teardown we have to wait until the teardown descriptor
is returned by the hardware. The g_zero testcase "testusb -a -t 9" triggers
the following warning quite reliable:
|------------[ cut here ]------------
|WARNING: CPU: 0 PID: 0 at drivers/dma/cppi41.c:609 cppi41_dma_control+0x198/0x304()
|[<c003f84c>] (warn_slowpath_null) from [<c02be8d8>]
|[<c02be8d8>] (cppi41_dma_control) from [<bf08d25c>]
|[<bf08d25c>] (cppi41_dma_channel_abort [musb_hdrc])
|[<bf08bc38>] (nuke.constprop.10 [musb_hdrc])
|[<bf08bd08>] (musb_gadget_disable [musb_hdrc])
|[<bf252524>] (disable_endpoints [usb_f_ss_lb])
|[<bf2525d8>] (disable_source_sink [usb_f_ss_lb])
|[<bf25260c>] (sourcesink_set_alt [usb_f_ss_lb])
|[<bf23ad24>] (composite_setup [libcomposite])
|[<bf08a2f4>] (musb_g_ep0_irq [musb_hdrc])
|[<bf085ec4>] (musb_interrupt [musb_hdrc])
|[<bf0aeaf4>] (dsps_interrupt [musb_dsps])
|[<c0080ea8>] (handle_irq_event_percpu)
|[<c008112c>] (handle_irq_event)
|[<c008348c>] (handle_level_irq)
|[<c00807a8>] (generic_handle_irq)
|[<c000ee80>] (handle_IRQ)
|[<c00085f0>] (omap3_intc_handle_irq)
and complains about a TD descriptor which is not returned. I've been
looking at several things and haven't noticed anything unusual that
might lead to this.
The manual says "to try again" until the descriptor comes out. I limited
the amount of retries to 100 retries in order to avoid an infinite number
of retries and so a busy-loop. Back then testing revealed that the
number of retries were around 20-30 so 100 seemed a good upper limit.
This g_zero test reaches without a problem 98 retries and it jumps
sometimes to 101 on am335x-evm and so the WARN_ON() triggers. Same test
run on beaglebone black and the retries start at 122 and my max value so
far was at 128.
So lets rise the limit to 500.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
There is no need to init .owner field.
Based on the patch from Peter Griffin <peter.griffin@linaro.org>
"mmc: remove .owner field for drivers using module_platform_driver"
This patch removes the superflous .owner field for drivers which
use the module_platform_driver API, as this is overriden in
platform_driver_register anyway."
Signed-off-by: Kiran Padwal <kiran.padwal@smartplayin.com>
[for nvidia]
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This change uses managed resource APIs to allocate resources such as,
mem, irq in order to simplify the driver unload or failure cases
Signed-off-by: Kiran Padwal <kiran.padwal@smartplayin.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
When a 0-length packet is received on the bus, desc->pd0 yields 1,
which confuses the driver's users. This information is clearly wrong
and not in accordance to the datasheet, but it's been observed on an
AM335x board, very reproducible.
Fix this by looking at bit 19 in PD2 of the completed packet. This bit
will tell us if a zero-length packet was received on a queue. If it's
set, ignore the value in PD0 and report a total length of 0 instead.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Start the channel tear down only if the channel is busy, else just
bail out. In some cases its seen that by the time the tear down is
initiated the cppi completes the DMA, especially in ISOCH transfers.
Signed-off-by: George Cherian <george.cherian@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Set the return variable to an error code as done elsewhere in the function.
A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
(
if@p1 (\(ret < 0\|ret != 0\))
{ ... return ret; }
|
ret@p1 = 0
)
... when != ret = e1
when != &ret
*if(...)
{
... when != ret = e2
when forall
return ret;
}
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Return code of pm_runtime_get_sync() > 0 is not an error and may happen.
Noticed during rmmod & modprobe testing.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Most of the logic here is try and error since what actually happens does
not match the trm or I miss read it.
My first assumption was that the queue on which the tear-down descriptor
completes (their own complete queue vs "active descriptor" complete
queue) depends on the transfer direction. This seems not to be true
because I manage to trigger
| WARN_ON(c->desc_phys != desc_phys);
and the other few were fine means the tear-down descriptor was valid but
on different queue.
This patch changes the logic here to look on both queues for the
descriptor.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Use cppi41_pop_desc() when appropriate instead of open-coding the same
functionality again. That makes the code more readable. The function has
to be moved some lines up for this change.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
With active users over suspend/resume cycles, it turns out that
more registers, in particular DMA_TDFDQ and RXHPCRA0, have to be
restored on resume.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This patch adds support for suspend/resume functionality to the cppi41
DMA driver. The steps necessary to make the system resume properly were
figured out by trial-and-error. The code as it stands now is the
minimum that has to be done to put the musb host system on an AM33xx
system into an operable state after resume.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Makes the code more readable and compact. No functional change.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Instead of passing around struct plafform_device, use struct device and
save one level of dereferencing. This affects the following functions:
* cppi41_add_chans
* purge_descs
* deinit_cpii41
* init_descs
* init_cppi41
* cppi_glue_infos
It's just a cosmetic cleanup that makes the code more readable.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
The test here should be ">=" instead of ">". The cdd->chan_busy[] array
has "ALLOC_DECS_NUM" elements.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With enabled pm_runtime in the kernel the device won't work because it
is not "on" during the probe function. This patch enables the device via
pm_runtime on probe so it remains activated.
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Before Randy figures out that this does not compile with CONFIG_BUG=n
here is a fix for it.
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
This driver is currently used by musb' cppi41 couter part. I may merge
both dma engine user of musb at some point but not just yet.
The driver seems to work in RX/TX mode in host mode, tested on mass
storage. I increaed the size of the TX / RX transfers and waited for the
core code to cancel a transfers and it seems to recover.
v2..3:
- use mall transfers on RX side and check data toggle.
- use rndis mode on tx side so we haveon interrupt for 4096 transfers.
- remove custom "transferred" hack and use dmaengine_tx_status() to
compute the total amount of data that has been transferred.
- cancel transfers and reclaim descriptors
v1..v2:
- RX path added
- dma mode 0 & 1 is working
- device tree nodes re-created.
Cc: Vinod Koul <vinod.koul@intel.com>
Cc: Dan Williams <djbw@fb.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
