Re: [RESEND PATCH V4 0/4] Introduce CoreSight STM support

From: Mathieu Poirier
Date: Fri Mar 25 2016 - 08:51:54 EST


On 7 March 2016 at 23:33, Chunyan Zhang <zhang.chunyan@xxxxxxxxxx> wrote:
> This patchset adds support for the CoreSight STM IP block.
>
> In the fourth version, comments from various people have been
> addressed. Representing configurations where channels are shared
> between multiple masterIDs has been kept unchanged from the previous
> version because a viable alternative hasn't been suggested.
>
> This RESEND PATCH 1/4 depends on the patch [3] which has been
> merged into linux-next.

We are not going anywhere with 1/4. Since the functionality conveyed
by that patch isn't strictly needed simply drop it (along with 2/4).
Please take into consideration the comments I made on 4/4 and respin
on 4.6-rc1 when it comes out.

Thanks,
Mathieu

>
> Changes from V3:
> - Removed ioctl get_options interface from the generic STM code and CoreSight STM driver.
> - Removed 'write_max' from the structure 'stm_drvdata', and changed 'write_64bit' to 'write_bytes'.
> - Revised stm_fundamental_data_size() to return the fundamental data size instead of 0/1.
> - Removed stm_remove() from the driver.
> - Revised the return value of ::packet() callback function according to [2].
> - Modified stm_send() to send one STP packet at a time.
> - Added comments to invariant/guaranteed CoreSight STM transaction mode.
>
> Changes from V2:
> - Changed to return -EFAULT if failed on the command STP_GET_OPTIONS.
> - Used Alex's patch [1] instead of the last 2/6.
> - Removed the while loop from stm_send(), since the packet size passed
> to it isn't larger than 4 bytes on 32-bit system and 8 bytes on
> 64-bit system.
> - Removed stm_send_64bit(), since the process of packets on 64-bit
> CS-STM should be basically the same with on 32-bit system, except the
> maximum length of writing STM at a time.
> - Removed the support of writing 64-bit to CoreSight STM buffer at a time
> on 32-bit ARM architecture according to an ARM engineer suggestion. As
> he said that the STM might receive a 64-bit write, or might receive a
> pair of 32-bit writes to the two addressed words in either order.
> So 64-bit write isn't guaranteed to work on the ARM 32-bit architecture.
>
> Changes from v1:
> - Added a definition of coresight_simple_func() in CS-STM driver to
> avoid the kbuild test robot error for the time being. This
> modification will be removed when merging the code in which the
> coresight_simple_func() has been moved to the header file.
> - Calculate the channel number according to the channel memory space size.
>
>
> Thanks,
> Chunyan
>
> [1] https://lkml.org/lkml/2016/2/4/652
> [2] https://lkml.org/lkml/2016/2/12/397
> [3] https://lkml.org/lkml/2015/12/22/348
>
> Chunyan Zhang (1):
> Documentations: Add explanations of the case for non-configurable
> masters
>
> Mathieu Poirier (2):
> stm class: provision for statically assigned masterIDs
> coresight-stm: Bindings for System Trace Macrocell
>
> Pratik Patel (1):
> coresight-stm: adding driver for CoreSight STM component
>
> .../ABI/testing/sysfs-bus-coresight-devices-stm | 53 ++
> .../devicetree/bindings/arm/coresight.txt | 28 +
> Documentation/trace/coresight.txt | 37 +-
> Documentation/trace/stm.txt | 6 +
> drivers/hwtracing/coresight/Kconfig | 11 +
> drivers/hwtracing/coresight/Makefile | 1 +
> drivers/hwtracing/coresight/coresight-stm.c | 890 +++++++++++++++++++++
> drivers/hwtracing/stm/core.c | 17 +-
> drivers/hwtracing/stm/policy.c | 18 +-
> include/linux/coresight-stm.h | 6 +
> include/linux/stm.h | 8 +
> include/uapi/linux/coresight-stm.h | 21 +
> 12 files changed, 1090 insertions(+), 6 deletions(-)
> create mode 100644 Documentation/ABI/testing/sysfs-bus-coresight-devices-stm
> create mode 100644 drivers/hwtracing/coresight/coresight-stm.c
> create mode 100644 include/linux/coresight-stm.h
> create mode 100644 include/uapi/linux/coresight-stm.h
>
> --
> 1.9.1
>