[RFC][WIP]Nhlt plugin and simple topology plugin interface #5

juimonen · 2021-12-15T11:24:54Z

No description provided.

juimonen · 2021-12-15T11:27:20Z

Initial something... I need to figure out:

where to specify which plugins to load/run
where to install the plugins

juimonen · 2021-12-16T09:32:00Z

@perexg FYI. if you have time to comment... sorry this is against sof not-up-to-date tree, so only 2 last commits are relevant.

Don't have so much experience on the .so loading code, this is taken from alsa-plugins mostly. So my questions like above:

where to install these plugins
how to get the to-be-loaded plugin names and where
is the "hidden-visibility" gcc flag ok to hide other than .so entrypoint

I've now hardcoded the nhlt plugin loading and can pretty much use the nhlt as in my previous version.

lgirdwood · 2021-12-21T17:29:59Z

@juimonen probably worth to send to alsa repo now.

Signed-off-by: Jaroslav Kysela <perex@perex.cz>

Introduce the -K,--lock-dir parameter to specify the locking directory. If the locking is active, files card<NUM>.lock are used to serialize the multiple initialization requests (udev, service). Allow the relative -O,--lock-file argument (it's default now). Signed-off-by: Jaroslav Kysela <perex@perex.cz>

topology/pre-processor.c

topology/pre-processor.h

topology/nhlt/intel/dmic-nhlt.c

plbossart · 2022-02-16T16:51:49Z

topology/nhlt/intel/dmic-nhlt.c

+/* get dmic endpoint count */
+int nhlt_dmic_get_ep_count(struct intel_nhlt_params *nhlt)
+{
+	return dmic_get_vendor_blob_count(nhlt);


is this correct? an endpoint can support multiple formats, and each format will expose a blob. The blob count is not identical to endpoint count IMHO.

so it is bad naming... this 1 "vendor blob" would include all hw config variants

topology/nhlt/intel/dmic-nhlt.c

plbossart · 2022-02-16T17:11:08Z

topology/nhlt/intel/ssp-nhlt.c

+	ep.revision_id = 0;
+	ep.subsystem_id = 0;
+	ep.device_type = 0;
+	ep.direction = NHLT_ENDPOINT_DIRECTION_RENDER;


what about capture?

so not sure what to do here... NHLT doesn't have duplex definition, so should I just duplicate the endpoint data in that case? anyway can take this somehow from direction param

Yes I think the endpoints are duplicated for the SSP. That's what you see from the NHTL tool from Bob Moore.

sure I can add the duplication

plbossart · 2022-02-16T17:13:41Z

topology/nhlt/intel/ssp-nhlt.c

+ *			id	0
+ *			mclk_freq	38400000
+ *			bclk_freq	4800000
+ *			tdm_slot_width	32


this example cannot work, the bclk would be 3072000 Hz...

oh, maybe bclk 2400000

nope, blck of 2.4 MHz cannot work because that would mean slots of 25 bits that NHLT cannot represent.

for NHLT, you have to assume the slot_width is a multiple of 8, and then find the bclk=fsync * number of slots * slots_width.

that will not work of course for platforms with a 24 MHz root clock, where I think the only solution might be to use the dreaded M/N dividers.

plbossart · 2022-02-16T17:14:55Z

topology/nhlt/intel/ssp/ssp-intel.h

+	uint32_t ssc3;
+	uint32_t ssioc;
+	uint32_t mdivc;
+	uint32_t mdivr;


hmm I have here mdivc and mdivr separately, I see them in some spec mdiv[2], is that what you mean?

yes, we have two MCLK IDs, and you can configure them separately.

plbossart · 2022-02-16T17:16:35Z

topology/nhlt/intel/ssp/ssp-process.c

+			ssp_freq[MAX_SSP_FREQ_INDEX].freq);
+		return -EINVAL;
+	}
+#endif


hmm this is a check against maximum possible ssp frequency, which I understand is platform dependent. So not sure I can get this somewhere nicely.

for e.g. in tgl this would be 96000000, so if this is added to the topology I can do the check

but yeah will now just check for 0 mclk

plbossart · 2022-02-16T17:17:55Z

topology/nhlt/intel/ssp/ssp-process.c

+	else
+		clk_div = 0xFFF; /* bypass clk divider */
+
+	/* use clock source 0 */


why ? Some folks need MCLK1!

ok, I need to discuss this with you as this is a copy from initial code against closed fw...

Add a simple plugin interface for processing the topology tree. There can be cases where parts of the topology need to be translated from the original format into something else. For example one could calculate some kind of filter coefficients from filter parameters or some other binary interface parameters from plain text topology tokens. Mechanism is similar as in alsa-plugins and in the plugin there should be only 1 function exported of the form: int _snd_topology_##pluginname##_process (snd_config_t *input, snd_config_t *output) Input and output parameters are snd_config tree before and after topology2 pre-processing. So the plugin can modify both if needed. There are cases where the plugin may need to get information from input tree, but make modifications to the output. The plugins to be used can be defined in command line with: alsatplg -DPREPROCESS_PLUGINS="foobar" -c topology.conf -p -o topology.tplg Multiple plugins should be separated by ":". Plugins to be used can also be defined with "Define" clause inside the topology file (but command line takes precedence): Define { PREPROCESS_PLUGINS "foobar" } Signed-off-by: Jaska Uimonen <jaska.uimonen@linux.intel.com>

lgirdwood · 2022-02-21T10:37:06Z

@juimonen is this ready to send upstream once SSP is aligned (now DMIC is working) or are there other items WIP ?

Add Intel nhlt acpi table encoder plugin into topology2.0 processing. Nhlt internal structure is defined in: https://01.org/sites/default/files/595976_intel_sst_nhlt.pdf Nhlt acpi table contain vendor specific binary data blobs that are used in some Intel dsp platforms for configuring the dmic and ssp hardware. The function of this code is mainly to generate the vendor specific binary blobs, but as there is existing nhlt parser code and header in kernel there's no point of re-inventing the container: just use the existing nhlt acpi table format. Basically this code is creating similar nhlt acpi table that you would get from: cat /sys/firmware/acpi/tables/NHLT This code will have implementation for dmic and ssp endpoints. Thus the code will translate the topology dai tokens into vendor specific binary blobs and pack them into nhlt acpi table. Ssp and dmic code is lifted from Sound Open Firmware (sof) code base, thus it will have BSD-3 license. This plugin can be enabled from command line with: alsatplg -DPREPROCESS_PLUGINS="nhlt" -c foo.conf -p -o bar.tplg You can also dump the nhlt binary into a file with additional define: -DNHLT_BIN="nhlt.bin" Signed-off-by: Jaska Uimonen <jaska.uimonen@linux.intel.com>

alsactl distributed as part of Fedora 40 got a SEGV: # journalctl ... May 17 00:55:58 dev64.localdomain kernel: alsactl[1923]: segfault at 28 ip 00005600705b3373 sp 00007ffd9712bef0 error 4 in alsactl[5600705af000+13000] likely on CPU 5 (core 8, socket 0) ... As the following output of the debug session, card_free() tried a card pointing NULL: $ sudo coredumpctl debug alsactl PID: 1923 (alsactl) UID: 0 (root) GID: 0 (root) Signal: 11 (SEGV) Timestamp: Fri 2024-05-17 00:55:58 JST (3h 34min ago) Command Line: /usr/sbin/alsactl -s -n 19 -c -E ALSA_CONFIG_PATH=/etc/alsa/alsactl.conf --initfile=/lib/alsa/init/00main rdaemon Executable: /usr/sbin/alsactl Control Group: /system.slice/alsa-state.service Unit: alsa-state.service Slice: system.slice Boot ID: 241b5a2ef86f4940bb3d340583c80d88 Machine ID: 437365709a8c488c9481ee4b6651c2ec Hostname: dev64.localdomain Storage: /var/lib/systemd/coredump/core.alsactl.0.241b5a2ef86f4940bb3d340583c80d88.1923.1715874958000000.zst (present) Size on Disk: 81.7K Package: alsa-utils/1.2.11-1.fc40 build-id: 3b6fec58b3566d666d6e9fd48e8fcf04f03f0152 Message: Process 1923 (alsactl) of user 0 dumped core. Module libasound.so.2 from rpm alsa-lib-1.2.11-2.fc40.x86_64 Module alsactl from rpm alsa-utils-1.2.11-1.fc40.x86_64 Stack trace of thread 1923: #0 0x00005600705b3373 card_free (alsactl + 0xa373) #1 0x00005600705c0e54 state_daemon (alsactl + 0x17e54) #2 0x00005600705b2339 main (alsactl + 0x9339) #3 0x00007f4c0b9b7088 __libc_start_call_main (libc.so.6 + 0x2a088) #4 0x00007f4c0b9b714b __libc_start_main_impl (libc.so.6 + 0x2a14b) #5 0x00005600705b2df5 _start (alsactl + 0x9df5) ELF object binary architecture: AMD x86-64 GNU gdb (Fedora Linux) 14.2-1.fc40 Copyright (C) 2023 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> This is free software: you are free to change and redistribute it. There is NO WARRANTY, to the extent permitted by law. Type "show copying" and "show warranty" for details. This GDB was configured as "x86_64-redhat-linux-gnu". Type "show configuration" for configuration details. For bug reporting instructions, please see: <https://www.gnu.org/software/gdb/bugs/>. Find the GDB manual and other documentation resources online at: <http://www.gnu.org/software/gdb/documentation/>. For help, type "help". Type "apropos word" to search for commands related to "word"... Reading symbols from /usr/sbin/alsactl... Reading symbols from /usr/lib/debug/usr/sbin/alsactl-1.2.11-1.fc40.x86_64.debug... [New LWP 1923] [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib64/libthread_db.so.1". Core was generated by `/usr/sbin/alsactl -s -n 19 -c -E ALSA_CONFIG_PATH=/etc/alsa/alsactl.conf --init'. Program terminated with signal SIGSEGV, Segmentation fault. #0 free_list (list=0x20) at /usr/src/debug/alsa-utils-1.2.11-1.fc40.x86_64/alsactl/daemon.c:73 73 for (i = 0; i < list->size; i++) (gdb) where #0 free_list (list=0x20) at /usr/src/debug/alsa-utils-1.2.11-1.fc40.x86_64/alsactl/daemon.c:73 #1 card_free (card=card@entry=0x5600707455f0) at /usr/src/debug/alsa-utils-1.2.11-1.fc40.x86_64/alsactl/daemon.c:82 #2 0x00005600705c0e54 in state_daemon (file=file@entry=0x5600705c31a1 "/var/lib/alsa/asound.state", cardname=cardname@entry=0x0, period=period@entry=300, pidfile=pidfile@entry=0x5600705c3170 "/var/run/alsactl.pid") at /usr/src/debug/alsa-utils-1.2.11-1.fc40.x86_64/alsactl/daemon.c:455 #3 0x00005600705b2339 in main (argc=<optimized out>, argv=<optimized out>) at /usr/src/debug/alsa-utils-1.2.11-1.fc40.x86_64/alsactl/alsactl.c:459 (gdb) list 68 69 static void free_list(struct id_list *list) 70 { 71 int i; 72 73 for (i = 0; i < list->size; i++) 74 free(list->list[i]); 75 free(list->list); 76 } 77 (gdb) up #1 card_free (card=card@entry=0x5600707455f0) at /usr/src/debug/alsa-utils-1.2.11-1.fc40.x86_64/alsactl/daemon.c:82 82 free_list(&c->blacklist); (gdb) p c $1 = (struct card *) 0x0 (gdb) Closes: alsa-project#267 Signed-off-by: Masatake YAMATO <yamato@redhat.com> Signed-off-by: Jaroslav Kysela <perex@perex.cz>

juimonen requested a review from ranj063 December 15, 2021 11:27

juimonen force-pushed the nhlt_plugin branch from 925309d to 6293c61 Compare December 15, 2021 12:58

juimonen force-pushed the nhlt_plugin branch from 6293c61 to d75c3eb Compare December 21, 2021 19:52

juimonen force-pushed the nhlt_plugin branch 2 times, most recently from 97f9e79 to 5523cd4 Compare January 14, 2022 13:16

perexg added 2 commits February 1, 2022 19:47

alsactl: add dbg() prints to init_ucm()

f5997e1

Signed-off-by: Jaroslav Kysela <perex@perex.cz>

alsactl: remove unused argument for state_lock_()

ccb6d46

Signed-off-by: Jaroslav Kysela <perex@perex.cz>

juimonen force-pushed the nhlt_plugin branch from 5523cd4 to 5e57ae3 Compare February 1, 2022 19:32

alsactl: lock - use alarm signal and F_SETLKW rather polling

64fa37f

Signed-off-by: Jaroslav Kysela <perex@perex.cz>

juimonen force-pushed the nhlt_plugin branch from 5e57ae3 to 780a540 Compare February 1, 2022 20:48

juimonen force-pushed the nhlt_plugin branch from 780a540 to 0e617d4 Compare February 16, 2022 15:32

plbossart reviewed Feb 16, 2022

View reviewed changes

juimonen force-pushed the nhlt_plugin branch 4 times, most recently from bac507f to a6ea097 Compare February 17, 2022 14:59

juimonen force-pushed the nhlt_plugin branch from a6ea097 to 77065c1 Compare February 18, 2022 15:31

juimonen force-pushed the nhlt_plugin branch 2 times, most recently from 0382c29 to 551df49 Compare February 28, 2022 16:50

juimonen force-pushed the nhlt_plugin branch 4 times, most recently from 7f28666 to b76bdc7 Compare March 4, 2022 14:34

juimonen force-pushed the nhlt_plugin branch from b76bdc7 to dc3bbb1 Compare March 9, 2022 12:16

juimonen closed this Mar 15, 2022

[RFC][WIP]Nhlt plugin and simple topology plugin interface #5

[RFC][WIP]Nhlt plugin and simple topology plugin interface #5

Uh oh!

Conversation

juimonen commented Dec 15, 2021

Uh oh!

juimonen commented Dec 15, 2021

Uh oh!

juimonen commented Dec 16, 2021

Uh oh!

lgirdwood commented Dec 21, 2021

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lgirdwood commented Feb 21, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants