espressif
ESP32C6: regdma_link_get_next hit Guru Meditation Error: Core 0 panic'ed (Stack protection fault) in master (IDFGH-12238)
Answers checklist.
- [X] I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
- [X] I have updated my IDF branch (master or release) to the latest version and checked that the issue is present there.
- [X] I have searched the issue tracker for a similar issue and not found a similar issue.
IDF version.
v5.3-dev-2218-g7cf8cc79a1
Espressif SoC revision.
ESP32C6
Operating System used.
Linux
How did you build your project?
Command line with idf.py
If you are using Windows, please specify command line type.
None
Development Kit.
Custom Board
Power Supply used.
USB
What is the expected behavior?
no panic, as exactly the same application code works in v5.2 branch.
What is the actual behavior?
Guru Meditation Error: Core 0 panic'ed (Stack protection fault).
Steps to reproduce.
My application is running nimble host-only using uart hci but I cannot share my application code.
Debug Logs.
Guru Meditation Error: Core 0 panic'ed (Stack protection fault).
Detected in task "nimble_host" at 0x4200471c
0x4200471c: regdma_link_get_next at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:351
Stack pointer: 0x4084de90
Stack bounds: 0x4084dea8 - 0x4084eea0
Stack dump detected
Core 0 register dump:
MEPC : 0x42004750 RA : 0x420047cc SP : 0x4084de90 GP : 0x40818354
0x42004750: regdma_link_get_next at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:363
0x420047cc: regdma_link_recursive_impl at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390 (discriminator 1)
(inlined by) regdma_link_recursive_impl at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385 (discriminator 1)
TP : 0x40818a7c T0 : 0x40022494 T1 : 0x4080288a T2 : 0xffffffff
0x40022494: multi_heap_internal_unlock in ROM
0x4080288a: uart_ll_disable_intr_mask at /home/axel/esp/esp-idf/components/hal/esp32c6/include/hal/uart_ll.h:445
(inlined by) uart_rx_intr_handler_default at /home/axel/esp/esp-idf/components/esp_driver_uart/src/uart.c:914
S0/FP : 0x40828620 S1 : 0x00000003 A0 : 0x40828620 A1 : 0x00000000
A2 : 0x00000052 A3 : 0x420048d0 A4 : 0x70030000 A5 : 0x4210b000
0x420048d0: regdma_link_update_stats_wrapper at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:442
A6 : 0x00000004 A7 : 0x0000000a S2 : 0x00000000 S3 : 0x00000052
S4 : 0x00000004 S5 : 0x00000006 S6 : 0x00000000 S7 : 0x00000000
S8 : 0x00000003 S9 : 0x00000004 S10 : 0x42c80000 S11 : 0x41200000
T3 : 0x00000000 T4 : 0x00000001 T5 : 0x851eb850 T6 : 0x00000001
MSTATUS : 0x00001881 MTVEC : 0x40800001 MCAUSE : 0x0000001b MTVAL : 0xc8078793
0x40800001: _vector_table at ??:?
MHARTID : 0x00000000
Backtrace:
0x42004750 in regdma_link_get_next (link=link@entry=0x40828620, entry=entry@entry=0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:363
363 regdma_entry_buf_t *next = (*nextfn1[head.mode])(link);
#0 0x42004750 in regdma_link_get_next (link=link@entry=0x40828620, entry=entry@entry=0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:363
#1 0x420047cc in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=82, entry=0, link=0x40828620) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#2 regdma_link_recursive_impl (link=0x40828620, entry=0, depth=82, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#3 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=81, entry=0, link=0x4082864c) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#4 regdma_link_recursive_impl (link=0x4082864c, entry=0, depth=81, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#5 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=80, entry=0, link=0x40828678) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#6 regdma_link_recursive_impl (link=0x40828678, entry=0, depth=80, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#7 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=79, entry=0, link=0x40828714) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#8 regdma_link_recursive_impl (link=0x40828714, entry=0, depth=79, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#9 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=78, entry=0, link=0x40828738) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#10 regdma_link_recursive_impl (link=0x40828738, entry=0, depth=78, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#11 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=77, entry=0, link=0x408287fc) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#12 regdma_link_recursive_impl (link=0x408287fc, entry=0, depth=77, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#13 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=76, entry=0, link=0x408288f8) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#14 regdma_link_recursive_impl (link=0x408288f8, entry=0, depth=76, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#15 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=75, entry=0, link=0x40828994) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#16 regdma_link_recursive_impl (link=0x40828994, entry=0, depth=75, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#17 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=74, entry=0, link=0x408289b8) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#18 regdma_link_recursive_impl (link=0x408289b8, entry=0, depth=74, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#19 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=73, entry=0, link=0x40828a7c) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#20 regdma_link_recursive_impl (link=0x40828a7c, entry=0, depth=73, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#21 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=72, entry=0, link=0x40828b78) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#22 regdma_link_recursive_impl (link=0x40828b78, entry=0, depth=72, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#23 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=71, entry=0, link=0x40828ba4) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#24 regdma_link_recursive_impl (link=0x40828ba4, entry=0, depth=71, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#25 0x420047d6 in regdma_link_recursive_impl (hook=0x420048d0 <regdma_link_update_stats_wrapper>, depth=70, entry=0, link=0x40828bd0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#26 regdma_link_recursive_impl (link=0x40828bd0, entry=0, depth=70, hook=0x420048d0 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#27 0x00000000 in ?? ()
Backtrace stopped: frame did not save the PC
ELF file SHA256: ed9c36e055e3fa76
Rebooting...
ESP-ROM:esp32c6-20220919
More Information.
No response
ff94741cd13fbd43a3805ead2b60cc59ebb8f212 is good commit, 2302dd5a91cac6dce2853feda780a1b90a056001 is bad commit.
So it looks like below commit causes this regression: Merge branch 'feature/i2c_sleep' into 'master'
feat(i2c): Support i2c sleep retention on esp32c6/h2
@AxelLin Please try to increase the size of CONFIG_BT_NIMBLE_HOST_TASK_STACK_SIZE in menuconfig.
@AxelLin Please try to increase the size of
CONFIG_BT_NIMBLE_HOST_TASK_STACK_SIZEin menuconfig.
Why? The error is not stack overflow. Anyway , tested CONFIG_BT_NIMBLE_HOST_TASK_STACK_SIZE=8192 still hit the same error.
Guru Meditation Error: Core 0 panic'ed (Stack protection fault).
It is indeed caused by stack overflow, refer: https://docs.espressif.com/projects/esp-idf/en/v5.2/esp32c6/api-guides/fatal-errors.html#stack-overflow
I'm guessing it's caused by the fact that with i2c sleep retention support, the depth of function calls in the sleep retention de-initialize flow has become deeper.
@esp-wzh It's in reviewing state, I'm wondering if you can provide the patch for testing.
@AxelLin
Stack pointer: 0x4084de90
Stack bounds: 0x4084dea8 - 0x4084eea0
I'm not sure how to reproduce the problem you're having, from the log attached, task nimble_host has a stack region of 0x4084dea8 - 0x4084eea0, but during the task run the Stack Pointer crosses over the minimum address.
Can you share the log when the stack overflowed after you increased the size of CONFIG_BT_NIMBLE_HOST_TASK_STACK_SIZE to 8192 Byte ?
Or can you help provide a minimum reproduction example?
I'm guessing it's caused by the fact that with i2c sleep retention support, the depth of function calls in the sleep retention de-initialize flow has become deeper.
@esp-wzh
I tested again and now I notice the stackover flow could happen on various tasks. This is a problem, which means I need to increase stacksize for various tasks.
Below is an example of the issue happen on "main" task.
Detected in task "main" at 0x420049e2
0x420049e2: regdma_link_get_next at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:351
Stack pointer: 0x4082b080
Stack bounds: 0x4082b090 - 0x4082c280
Stack dump detected
Core 0 register dump:
MEPC : 0x42004a84 RA : 0x42004ade SP : 0x4082b080 GP : 0x40819b44
0x42004a84: regdma_link_get_next at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:376
0x42004ade: regdma_link_recursive_impl at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390 (discriminator 1)
(inlined by) regdma_link_recursive_impl at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385 (discriminator 1)
TP : 0x4082c280 T0 : 0x40022494 T1 : 0x40802aa0 T2 : 0xffffffff
0x40022494: multi_heap_internal_unlock in ROM
0x40802aa0: uart_ll_disable_intr_mask at /home/axel/esp/esp-idf/components/hal/esp32c6/include/hal/uart_ll.h:445
(inlined by) uart_rx_intr_handler_default at /home/axel/esp/esp-idf/components/esp_driver_uart/src/uart.c:914
S0/FP : 0x40835fa8 S1 : 0x42004c12 A0 : 0x40835fa8 A1 : 0x00000000
0x42004c12: regdma_link_update_stats_wrapper at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:442
A2 : 0x00000067 A3 : 0x42004c12 A4 : 0x00000003 A5 : 0x00000003
0x42004c12: regdma_link_update_stats_wrapper at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:442
A6 : 0x00000004 A7 : 0x0000000a S2 : 0x00000000 S3 : 0x00000067
S4 : 0x00000004 S5 : 0x00000006 S6 : 0x00000000 S7 : 0x00000000
S8 : 0x40825000 S9 : 0x40825000 S10 : 0x00000000 S11 : 0x00000000
T3 : 0x00000000 T4 : 0x00000001 T5 : 0x40825000 T6 : 0x00000001
MSTATUS : 0x00001881 MTVEC : 0x40800001 MCAUSE : 0x0000001b MTVAL : 0x00004732
0x40800001: _vector_table at ??:?
MHARTID : 0x00000000
Backtrace:
regdma_link_get_next (link=link@entry=0x40835fa8, entry=entry@entry=0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:376
376 void *next = (*nextfn0[head.mode])(link);
#0 regdma_link_get_next (link=link@entry=0x40835fa8, entry=entry@entry=0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:376
#1 0x42004ade in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=103, entry=0, link=0x40835fa8) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#2 regdma_link_recursive_impl (link=0x40835fa8, entry=0, depth=103, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#3 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=102, entry=0, link=0x408360ac) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#4 regdma_link_recursive_impl (link=0x408360ac, entry=0, depth=102, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#5 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=101, entry=0, link=0x408360cc) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#6 regdma_link_recursive_impl (link=0x408360cc, entry=0, depth=101, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#7 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=100, entry=0, link=0x408360ec) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#8 regdma_link_recursive_impl (link=0x408360ec, entry=0, depth=100, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#9 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=99, entry=0, link=0x4083610c) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#10 regdma_link_recursive_impl (link=0x4083610c, entry=0, depth=99, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#11 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=98, entry=0, link=0x40836184) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#12 regdma_link_recursive_impl (link=0x40836184, entry=0, depth=98, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#13 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=97, entry=0, link=0x40829fac) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#14 regdma_link_recursive_impl (link=0x40829fac, entry=0, depth=97, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#15 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=96, entry=0, link=0x40829fd8) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#16 regdma_link_recursive_impl (link=0x40829fd8, entry=0, depth=96, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#17 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=95, entry=0, link=0x40829ff8) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#18 regdma_link_recursive_impl (link=0x40829ff8, entry=0, depth=95, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#19 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=94, entry=0, link=0x4082a020) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#20 regdma_link_recursive_impl (link=0x4082a020, entry=0, depth=94, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#21 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=93, entry=0, link=0x4082a040) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#22 regdma_link_recursive_impl (link=0x4082a040, entry=0, depth=93, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#23 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=92, entry=0, link=0x4082a060) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#24 regdma_link_recursive_impl (link=0x4082a060, entry=0, depth=92, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#25 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=91, entry=0, link=0x4082a080) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#26 regdma_link_recursive_impl (link=0x4082a080, entry=0, depth=91, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#27 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=90, entry=0, link=0x4082a0a0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#28 regdma_link_recursive_impl (link=0x4082a0a0, entry=0, depth=90, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#29 0x42004ae8 in regdma_link_recursive_impl (hook=0x42004c12 <regdma_link_update_stats_wrapper>, depth=89, entry=0, link=0x4082a1c0) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:390
#30 regdma_link_recursive_impl (link=0x4082a1c0, entry=0, depth=89, hook=0x42004c12 <regdma_link_update_stats_wrapper>) at /home/axel/esp/esp-idf/components/esp_hw_support/port/regdma_link.c:385
#31 0x00000000 in ?? ()
Backtrace stopped: frame did not save the PC
ELF file SHA256: 11864d7830376559
Rebooting...
ESP-ROM:esp32c6-20220919
Build:Sep 19 2022
rst:0xc (SW_CPU),boot:0x6c (SPI_FAST_FLASH_BOOT)
@esp-wzh
FYI, see the HWM changes:
v5.1 branch: Task Name Status Prio HWM Task# nimble_host B 21 5592 10
master: Task Name Status Prio HWM Task# nimble_host B 21 3176 10
master (ff94741, before supporting i2c sleep retention on esp32c6/h2) Task Name Status Prio HWM Task# nimble_host B 21 5608 10
Can you briefly describe what you did in the main task and how it called into function regdma_link?
Usually regdma_link_recursive_impl will be called when module deinit destroys its retention link. as you can see in the backtrace, we currently use a recursive algorithm to search for nodes that need to be destroyed in the linked list.
I have encountered this problem before, and realized that this algorithm is not suitable for application in embedded systems with scarce RAM. We are exploring better solutions, which are already in our TODO List.
As analyzed above, if the module de-initialization(bt, i2c, gdma ...) is called by a fixed task, you can temporarily avoid this problem by increasing the stack size of the corresponding task.
@esp-wzh
FYI, see the HWM changes:
v5.1 branch: Task Name Status Prio HWM Task# nimble_host B 21 5592 10
master: Task Name Status Prio HWM Task# nimble_host B 21 3176 10
master (ff94741, before supporting i2c sleep retention on esp32c6/h2) Task Name Status Prio HWM Task# nimble_host B 21 5608 10
It makes sense because after i2c retention is supported, destroying the link requires traversing more nodes, going through deeper recursion, and using more of the stack.
I made a simple attempt. Adding the FORCE_INLINE_ATTR attribute to the regdma_link_get_next can save about 25% of the call stack consumption. It may be helpful to you.
@esp-wzh FYI, see the HWM changes: v5.1 branch: Task Name Status Prio HWM Task# nimble_host B 21 5592 10 master: Task Name Status Prio HWM Task# nimble_host B 21 3176 10 master (ff94741, before supporting i2c sleep retention on esp32c6/h2) Task Name Status Prio HWM Task# nimble_host B 21 5608 10
It makes sense because after i2c retention is supported, destroying the link requires traversing more nodes, going through deeper recursion, and using more of the stack.
It takes 2KB+ stack size in current master (compare to master (ff94741)). I don't think this is ok, which is likely to cause regression for existing application.
@esp-wzh
Just remind this issue can happen on different tasks. Please fix this, otherwise the master tree is simply broken. It's impossible to increase 2KB+ stack size for each task.
@esp-wzh
I'm still waiting for this fix, any update? If it really needs much longer time to fix it, I think you should revert the change.
Can you share a minimal demo that reproduces this issue? @AxelLin
Sigh, which means no progress in the past month. too bad. You are aware the issue is related to stack size, which means you can check task HWM to understand why it causes issue. (https://github.com/espressif/esp-idf/issues/13288#issuecomment-1976081673) I believe you can easily reproduce the issue with adjusting the stack size.
@AxelLin I added the method of using iterators to optimize the stack consumption caused by recursion. The following figure shows the comparison before and after optimization (refer to the attachments regdma link recursive.log and regdma link iterator.log for detailed log files), and the patch file of optimization implementation refer to 0001-fix-esp_hw_support-fixed-stack-protection-fault-caus.patch.
@esp-lis @esp-wzh
Yes, the patch fixes the panic. Above log already shows the big difference about stack consumption: 3.2KB v.s. 48 bytes.
Please fix it ASAP, since this is a regression. BTW, the v5.2 branch (v5.2.1-443-g2ba5320112) also needs fix now.
@esp-lis @esp-wzh As I said this regression is now in v5.2 release branch, so when will this get fixed?
@esp-lis @esp-wzh As I said this regression is now in v5.2 release branch, so when will this get fixed?
The fix is not yet available in v5.2. When will you backport the fix. (I have been wait for more than 1 Month)