On Wed, 20 Nov 2024 at 09:39, Arnaud POULIQUEN arnaud.pouliquen@foss.st.com wrote:
On 11/20/24 17:04, Mathieu Poirier wrote:
On Tue, 19 Nov 2024 at 13:38, Mathieu Poirier mathieu.poirier@linaro.org wrote:
On Tue, 19 Nov 2024 at 11:14, Arnaud POULIQUEN arnaud.pouliquen@foss.st.com wrote:
Hello Mathieu,
On 11/18/24 18:52, Mathieu Poirier wrote:
On Mon, Nov 04, 2024 at 02:35:12PM +0100, Arnaud Pouliquen wrote:
This patch updates the rproc_ops struct to include an optional release_fw function.
The release_fw ops is responsible for releasing the remote processor firmware image. The ops is called in the following cases:
- An error occurs in rproc_start() between the loading of the segments and the start of the remote processor.
- after stopping the remote processor.
Signed-off-by: Arnaud Pouliquen arnaud.pouliquen@foss.st.com
Updates from version V11:
- fix typo in @release_fw comment
drivers/remoteproc/remoteproc_core.c | 5 +++++ include/linux/remoteproc.h | 3 +++ 2 files changed, 8 insertions(+)
diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c index 7694817f25d4..46863e1ca307 100644 --- a/drivers/remoteproc/remoteproc_core.c +++ b/drivers/remoteproc/remoteproc_core.c @@ -1258,6 +1258,9 @@ static int rproc_alloc_registered_carveouts(struct rproc *rproc)
static void rproc_release_fw(struct rproc *rproc) {
- if (rproc->ops->release_fw)
rproc->ops->release_fw(rproc);- /* Free the copy of the resource table */ kfree(rproc->cached_table); rproc->cached_table = NULL;
@@ -1377,6 +1380,8 @@ static int rproc_start(struct rproc *rproc, const struct firmware *fw) unprepare_subdevices: rproc_unprepare_subdevices(rproc); reset_table_ptr:
- if (rproc->ops->release_fw)
rproc->ops->release_fw(rproc);I suggest the following:
- Create two new functions, i.e rproc_load_fw() and rproc_release_fw(). The
only thing those would do is call rproc->ops->load_fw() and rproc->ops->release_fw(), if they are present. When a TEE interface is available, ->load_fw() and ->release_fw() become rproc_tee_load_fw() and rproc_tee_release_fw().
I'm wondering if it should be ->preload_fw() instead of ->load_fw() ops, as the ->load() op already exists.
I agree that ->load() and ->load_fw() will lead to confusion. I would support ->preload_fw() but there is no obvious antonyme.
Since we already have rproc_ops::prepare() and rproc_prepare_device() I suggest rproc_ops::prepare_fw() and rproc_prepare_fw(). The corollary would be rproc_ops::unprepare_fw() and rproc_unprepare_fm(). That said, I'm open to other ideas should you be interested in finding other alternatives.
Actually... A better approach might to rename rproc::load to rproc::load_segments. That way we can use rproc::load_fw() and rproc_load_fw() without confusion.
Concerning this proposal, please correct me if I'm wrong
- ops::load_segments() would be used for ELF format only as segment notion seems
linked to this format.
Correct - nothing different from what it is now.
- ops:rproc_load_fw should be used for other formats.
The risk is that someone may later come with a requirement to get a resource table first to configure some memories before loading a non-ELF firmware.
We can address that problem if/when it comes about.
- Call rproc_load_fw() in rproc_boot(), just before rproc_fw_boot(). If the
call to rproc_fw_boot() fails, call rproc_release_fw().
- The same logic applies to rproc_boot_recovery(), i.e call rproc_load_fw()
before rproc_start() and call rproc_release_fw() if rproc_start() fails.
I implemented this and I'm currently testing it. Thise second part requires a few adjustments to work. The ->load() ops needs to becomes optional to not be called if the "->preload_fw()" is used.
For that, I propose to return 0 in rproc_load_segments if rproc->ops->load is NULL and compensate by checking that at least "->preload_fw()" or ->load() is non-null in rproc_alloc_ops.
I agree.
Thanks, Arnaud
- Take rproc_tee_load_fw() out of rproc_tee_parse_fw(). It will now be called
in rproc_load_fw().
- As stated above function rproc_release_fw() now calls rproc_tee_release_fw().
The former is already called in rproc_shutdown() so we are good in that front.
With the above the cached_table management within the core remains the same and we can get rid of patch 3.7.
Thanks, Mathieu
return ret;diff --git a/include/linux/remoteproc.h b/include/linux/remoteproc.h index 2e0ddcb2d792..08e0187a84d9 100644 --- a/include/linux/remoteproc.h +++ b/include/linux/remoteproc.h @@ -381,6 +381,8 @@ enum rsc_handling_status {
- @panic: optional callback to react to system panic, core will delay
panic at least the returned number of milliseconds- @coredump: collect firmware dump after the subsystem is shutdown
- @release_fw: optional function to release the firmware image from ROM memories.
*/
This function is called after stopping the remote processor or in case of an errorstruct rproc_ops { int (*prepare)(struct rproc *rproc); @@ -403,6 +405,7 @@ struct rproc_ops { u64 (*get_boot_addr)(struct rproc *rproc, const struct firmware *fw); unsigned long (*panic)(struct rproc *rproc); void (*coredump)(struct rproc *rproc);
- void (*release_fw)(struct rproc *rproc);
};
/**
2.25.1