Skip to content

[Do Not Merge] ipc4: invalidate host mailbox cache after IPC completion#6414

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
lyakh wants to merge 2 commits into from
Closed

[Do Not Merge] ipc4: invalidate host mailbox cache after IPC completion #6414

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4dd55fd
ipc4: invalidate host mailbox cache after IPC completion
lyakh Oct 13, 2022
82c7398
mixin-mixout: don't erase source-info data on reset
lyakh Oct 14, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 0 additions & 7 deletions src/audio/mixin_mixout.c
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -935,17 +935,10 @@ static int mixin_reset(struct comp_dev *dev)

static int mixout_reset(struct comp_dev *dev)
{
struct mixout_data *mixout_data;
struct mixed_data_info __sparse_cache *mixed_data_info;
struct list_item *blist;

comp_dbg(dev, "mixout_reset()");

mixout_data = comp_get_drvdata(dev);
mixed_data_info = mixed_data_info_acquire(mixout_data->mixed_data_info);
memset(mixed_data_info->source_info, 0, sizeof(mixed_data_info->source_info));
mixed_data_info_release(mixed_data_info);

if (dev->pipeline->source_comp->direction == SOF_IPC_STREAM_PLAYBACK) {
list_for_item(blist, &dev->bsource_list) {
struct comp_buffer *source;
Expand Down
3 changes: 3 additions & 0 deletions src/ipc/ipc4/handler.c
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -1133,4 +1133,7 @@ void ipc_cmd(struct ipc_cmd_hdr *_hdr)

ipc_msg_send(&msg_reply, data, true);
}

dcache_invalidate_region((__sparse_force void __sparse_cache *)MAILBOX_HOSTBOX_BASE,
MAILBOX_HOSTBOX_SIZE);

@kv2019ikv2019iOct 14, 2022

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@lyakh This seems to have notable impact in large test runs.

The one thing I don't fully understand is where are we writing to the mailbox. I only see reads from this memory, so not clear how can we have dirty data in the cache that would be written on top of whatever host writes to area.

ipc_msg_send() only writes to the registers.

@lyakhlyakhOct 14, 2022

Copy link
Copy Markdown
CollaboratorAuthor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@kv2019i right, if we aren't writing there (and we shouldn't) then the DSP shouldn't mark the cache as dirty and therefore shouldn't write it back speculatively. So, indeed, probably it would be enough to make sure not to write there. Unfortunately, we do have a couple of locations where we write there. Probes do that, but they aren't used in our CI tests, so that wouldn't be the reason for failures. But I think I found more locations where that happens: in lines like

	config->frame_fmt = in_frame_fmt;

in copier.c. I'm working on further patches to fix that. If that's the only location, then fixing it should make this commit superfluous. We also need to check that we don't write caches back ourselves, and so far I don't see such cases.

}