* [i2c-mxs] Broken DMA operations from within a module
@ 2013-02-13 12:00 Enrico Scholz
2013-02-13 12:14 ` Russell King - ARM Linux
0 siblings, 1 reply; 3+ messages in thread
From: Enrico Scholz @ 2013-02-13 12:00 UTC (permalink / raw)
To: linux-arm-kernel
Hi,
the i2c-mxs driver does not work when transmit buffers are located
within module memory.
E.g. when doing something like
struct i2c_msg xfer = {
.buf = "something",
...
};
ret = i2c_transfer(client->adapter, &xfer, 1);
random data will be sent because buf is not DMA mapable ("something" is
located in the .strtab section within module memory). MXS has a memory
layout like
vmalloc : 0xc8800000 - 0xff000000 ( 872 MB)
lowmem : 0xc0000000 - 0xc8000000 ( 128 MB)
modules : 0xbf000000 - 0xc0000000 ( 16 MB)
.text : 0xc0008000 - 0xc04b1e8c (4776 kB)
.init : 0xc04b2000 - 0xc04e61ac ( 209 kB)
.data : 0xc04e8000 - 0xc051b578 ( 206 kB)
.bss : 0xc051b578 - 0xc05eba44 ( 834 kB)
and is CONFIG_FLATMEM. Buffers above are e.g. at 0xbf0cdaf0.
You can verify the problem e.g. with the edt-ft5x06.c touch driver;
dumping 'sg->dma_address' in mxs_dma_prep_slave_sg() shows physical
addresses like 0x3f0cdaf0 for such strings, which are outside of valid
memory (0x40000000-0x7fffffff).
Does the i2c layer require DMA mapable buffers in its xfer structures
(e.g. like usb), or is the problem in the generic ARM/MXS layer or in
the mxs i2c driver?
Enrico
^ permalink raw reply [flat|nested] 3+ messages in thread
* [i2c-mxs] Broken DMA operations from within a module
2013-02-13 12:00 [i2c-mxs] Broken DMA operations from within a module Enrico Scholz
@ 2013-02-13 12:14 ` Russell King - ARM Linux
2013-02-13 12:28 ` Enrico Scholz
0 siblings, 1 reply; 3+ messages in thread
From: Russell King - ARM Linux @ 2013-02-13 12:14 UTC (permalink / raw)
To: linux-arm-kernel
On Wed, Feb 13, 2013 at 01:00:50PM +0100, Enrico Scholz wrote:
> E.g. when doing something like
>
> struct i2c_msg xfer = {
> .buf = "something",
> ...
> };
>
> ret = i2c_transfer(client->adapter, &xfer, 1);
>
> random data will be sent because buf is not DMA mapable ("something" is
> located in the .strtab section within module memory). MXS has a memory
> layout like
DMA buffers must be located in lowmem *always*. This means placing them
in the .data segment does not work. USB has been through this already
and it's well known there...
^ permalink raw reply [flat|nested] 3+ messages in thread
* [i2c-mxs] Broken DMA operations from within a module
2013-02-13 12:14 ` Russell King - ARM Linux
@ 2013-02-13 12:28 ` Enrico Scholz
0 siblings, 0 replies; 3+ messages in thread
From: Enrico Scholz @ 2013-02-13 12:28 UTC (permalink / raw)
To: linux-arm-kernel
Russell King - ARM Linux <linux@arm.linux.org.uk> writes:
>> .buf = "something",
>> ...
>> random data will be sent because buf is not DMA mapable ("something" is
>> located in the .strtab section within module memory). MXS has a memory
>> layout like
>
> DMA buffers must be located in lowmem *always*. This means placing them
> in the .data segment does not work. USB has been through this already
> and it's well known there...
Yes; I know about USB and this requirement is explicitly mentioned in
the 'struct urb' documentation. But I can not find something similar
for i2c_msg and lot of i2c drivers are working with non DMA mapable
buffers.
Enrico
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2013-02-13 12:28 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-02-13 12:00 [i2c-mxs] Broken DMA operations from within a module Enrico Scholz
2013-02-13 12:14 ` Russell King - ARM Linux
2013-02-13 12:28 ` Enrico Scholz
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox