sched/tls: Wrap up_info_size into tls_info_size by xiaoxiang781216 · Pull Request #6365 · apache/nuttx

xiaoxiang781216 · 2022-06-05T05:33:52Z

Summary

and Initialize the idle thread local storage by tls_init_info

Impact

code refactor

Testing

rv-virt:knsh64

masayuki2009 · 2022-06-05T10:57:34Z


In file included from pthread/pthread_create.c:39:
pthread/pthread_create.c: In function 'nx_pthread_create':
Error: pthread/pthread_create.c:294:29: error: invalid application of 'sizeof' to incomplete type 'struct tls_info_s'
  294 |                             up_tls_size() + attr->stacksize,
      |                             ^~~~~~~~~~~
make[1]: *** [Makefile:58: pthread_create.o] Error 1
In file included from task/task_init.c:34:
task/task_init.c: In function 'nxtask_init':
Error: task/task_init.c:134:29: error: invalid application of 'sizeof' to incomplete type 'struct tls_info_s'
  134 |                             up_tls_size() + stack_size,
      |                             ^~~~~~~~~~~
make[1]: *** [Makefile:58: task_init.o] Error 1
make[1]: Target 'libsched.a' not remade because of errors.
make: *** [tools/LibTargets.mk:59: sched/libsched.a] Error 2
make: Target 'all' not remade because of errors.

xiaoxiang781216 · 2022-06-05T14:21:09Z

@masayuki2009 fixed, please try again.

masayuki2009 · 2022-06-05T23:15:47Z

@masayuki2009 fixed, please try again.

@xiaoxiang781216
Hmm, the issue still happens.

$ qemu-6.2/build/riscv64-softmmu/qemu-system-riscv64 -semihosting -nographic -cpu rv64 -smp 8 -M virt -bios none -kernel ./nuttx
ABC
NuttShell (NSH) NuttX-3.6.1
nsh> uname -a
NuttX 3.6.1 3e4e5e0b82-dirty Jun  6 2022 08:11:36 risc-v rv-virt
nsh> hello
Hello, World!!
[    0.000000] riscv_exception: EXCEPTION: Store/AMO access fault. MCAUSE: 0000000000000007
[    0.000000] riscv_exception: PANIC!!! Exception = 0000000000000007
[    0.000000] up_assert: Assertion failed at file:common/riscv_exception.c line: 89 task: hello
[    0.000000] riscv_registerdump: EPC: 0000000080005c66
[    0.000000] riscv_registerdump: A0: 0000000000000001 A1: 0000000080208038 A2: 0000000000000019 A3: 0000000080206c38
[    0.000000] riscv_registerdump: A4: 0000000080208018 A5: 0000000080208058 A6: 0000000000000101 A7: 0000000000000000
[    0.000000] riscv_registerdump: T0: 00000000800039c6 T1: 0000000000000000 T2: 0000000000000000 T3: 0000000000000000
[    0.000000] riscv_registerdump: T4: 0000000000000000 T5: 0000000000000000 T6: 0000000000000000
[    0.000000] riscv_registerdump: S0: 0000000080208040 S1: 0000000080206c00 S2: ffffffffffffffe0 S3: fffffffffffffff8
[    0.000000] riscv_registerdump: S4: ffffffffffffffe0 S5: 0000000000000000 S6: 0000000000000000 S7: 0000000000000000
[    0.000000] riscv_registerdump: S8: 0000000000000000 S9: 0000000000000000 S10: 0000000000000000 S11: 0000000000000000
[    0.000000] riscv_registerdump: SP: 0000000080209d90 FP: 0000000080208040 TP: 0000000000000000 RA: 0000000080005b8a
[    0.000000] riscv_dumpstate: sp:     0000000080200830
[    0.000000] riscv_dumpstate: IRQ stack:
[    0.000000] riscv_dumpstate:   base: 00000000802000f0
[    0.000000] riscv_dumpstate:   size: 0000000000000800
[    0.000000] riscv_stackdump: 0000000080200820: 802000f0 00000000 80006ce6 00000000 00000059 00000000 80014ed8 00000000
[    0.000000] riscv_stackdump: 0000000080200840: deadbeef deadbeef 00000000 00000000 ffffffe0 ffffffff fffffff8 ffffffff
[    0.000000] riscv_stackdump: 0000000080200860: ffffffe0 ffffffff 80209c88 00000000 00000007 00000000 800043d6 00000000
[    0.000000] riscv_stackdump: 0000000080200880: 00000009 00000000 80001464 00000000 00000000 00000000 80201050 00000000
[    0.000000] riscv_stackdump: 00000000802008a0: 80209c88 00000000 80001bb8 00000000 c0003200 00000000 80000f98 00000000
[    0.000000] riscv_stackdump: 00000000802008c0: 80209d58 00000000 80206c00 00000000 00000007 00000000 80000dfe 00000000
[    0.000000] riscv_dumpstate: sp:     0000000080209d90
[    0.000000] riscv_dumpstate: User stack:
[    0.000000] riscv_dumpstate:   base: 00000000c0004050
[    0.000000] riscv_dumpstate:   size: 00000000000007e0
[    0.000000] riscv_dumpstate: Kernel stack:
[    0.000000] riscv_dumpstate:   base: 0000000080209260
[    0.000000] riscv_dumpstate:   size: 0000000000000c00
[    0.000000] riscv_dumpstate: Kernel Stack
[    0.000000] riscv_stackdump: 0000000080209d80: 00000000 00000000 00040120 00000000 fffffff8 ffffffff 80209120 00000000
[    0.000000] riscv_stackdump: 0000000080209da0: fffffff8 ffffffff 80007a7c 00000000 00000000 00000000 00000000 00000000
[    0.000000] riscv_stackdump: 0000000080209dc0: 802090c2 00000000 80208f00 00000000 802090a0 00000000 800127d8 00000000
[    0.000000] riscv_stackdump: 0000000080209de0: 80207ec0 00000000 80010cba 00000000 802090c2 00000000 802090a0 00000000
[    0.000000] riscv_stackdump: 0000000080209e00: 80208f00 00000000 800119b2 00000000 00000000 00000000 00000000 00000000
[    0.000000] riscv_stackdump: 0000000080209e20: 00000000 00000000 80011c84 00000000 00000000 00000000 800039d2 00000000
[    0.000000] riscv_stackdump: 0000000080209e40: 00000000 00000000 800015bc 00000000 c00014de 00000000 00040020 00000000
[    0.000000] riscv_showtasks:    PID    PRI      USED     STACK   FILLED    COMMAND
[    0.000000] riscv_showtasks:   ----   ----       776      2048    37.8%    irq
[    0.000000] riscv_dump_task:      0      0      1328      3024    43.9%    Idle Task
[    0.000000] riscv_dump_task:      1    100      2000      2000   100.0%!   /system/bin/init
[    0.000000] riscv_dump_task:      2    100       536      2016    26.5%    hello

xiaoxiang781216 · 2022-06-06T04:01:16Z

@masayuki2009 fixed, please try again.

@xiaoxiang781216 Hmm, the issue still happens.

$ qemu-6.2/build/riscv64-softmmu/qemu-system-riscv64 -semihosting -nographic -cpu rv64 -smp 8 -M virt -bios none -kernel ./nuttx
ABC
NuttShell (NSH) NuttX-3.6.1
nsh> uname -a
NuttX 3.6.1 3e4e5e0b82-dirty Jun  6 2022 08:11:36 risc-v rv-virt
nsh> hello
Hello, World!!
[    0.000000] riscv_stackdump: 0000000080209e40: 00000000 00000000 800015bc 00000000 c00014de 00000000 00040020 00000000
[    0.000000] riscv_showtasks:    PID    PRI      USED     STACK   FILLED    COMMAND
[    0.000000] riscv_showtasks:   ----   ----       776      2048    37.8%    irq
[    0.000000] riscv_dump_task:      0      0      1328      3024    43.9%    Idle Task
[    0.000000] riscv_dump_task:      1    100      2000      2000   100.0%!   /system/bin/init
[    0.000000] riscv_dump_task:      2    100       536      2016    26.5%    hello

init task restore the stack to the original size(2000), but the consumption is still higher than the capacity.
All patch in PR #6337 is the code refactor and the stack consumption should be same as before except:
48a0a2e

@masayuki2009 the follow command look like just generate the kernel binary:

./tools/configure.sh rv-virt:knsh64
make -j8

Could you tell me how to generate hello and init elf in the kernel mode?
BTW, is it right to set CONFIG_SYSTEM_NSH=y here?
https://github.com/apache/incubator-nuttx/blob/master/boards/risc-v/qemu-rv/rv-virt/configs/knsh64/defconfig#L93

masayuki2009 · 2022-06-06T04:16:40Z

Could you tell me how to generate hello and init elf in the kernel mode?

Please try the following commands.

$ cd nuttx                                                                                                                                                                                                                               
$ ./tools/configure.sh rv-virt:knsh64
$ make V=1 -j7                                                                                                                                                                                                                           
$ make export V=1                                                                                                                                                                                                                        
$ cd ../apps                                                                                                                                                                                                                             
$ ./tools/mkimport.sh -x ../nuttx/nuttx-export-*.zip                                                                                                                                                                                     
$ make import V=1                                                                                                                                                                                                                        
$ cd ../nuttx
$ qemu-system-riscv64 -semihosting -nographic -cpu rv64 -smp 8 -M virt -bios none -kernel ./nuttx

BTW, is it right to set CONFIG_SYSTEM_NSH=y here?

Yes.

xiaoxiang781216 · 2022-06-06T08:16:51Z

Ok, I will repo locally later.

xiaoxiang781216 · 2022-06-07T04:37:10Z

@masayuki2009 the crash is made by: 11313e1
but the change only contain the code refactor, I will compare the generated code to find the real difference.

xiaoxiang781216 · 2022-06-07T07:00:26Z

@masayuki2009 Fix is here: #6377
I retitle this PR to refactor tls code during the review. Please try both.

Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>

xiaoxiang781216 · 2022-06-08T03:51:01Z

@masayuki2009 how about this patch?

masayuki2009 · 2022-06-08T04:28:36Z

@masayuki2009 how about this patch?

Could you rebase this PR to the latest master?

xiaoxiang781216 · 2022-06-08T06:09:13Z

Done.

xiaoxiang781216 requested a review from masayuki2009 June 5, 2022 05:33

xiaoxiang781216 mentioned this pull request Jun 5, 2022

Refine the implementation of TLS #6337

Merged

xiaoxiang781216 force-pushed the tls branch from 75b2d5b to 8359223 Compare June 5, 2022 14:20

xiaoxiang781216 changed the title ~~Revert "sched/tls: Don't compensate the tls size to the stack size"~~ sched/tls: Compensate the tls size to the stack size Jun 5, 2022

xiaoxiang781216 force-pushed the tls branch 2 times, most recently from 08ddce0 to 3e4e5e0 Compare June 5, 2022 18:01

xiaoxiang781216 force-pushed the tls branch from 3e4e5e0 to a121fd6 Compare June 7, 2022 06:57

xiaoxiang781216 changed the title ~~sched/tls: Compensate the tls size to the stack size~~ sched/tls: Wrap up_info_size into tls_info_size Jun 7, 2022

xiaoxiang781216 added 2 commits June 8, 2022 04:52

sched/init: Initialize the idle thread local storage by tls_init_info

6f212f8

Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>

sched/tls: Wrap up_info_size into tls_info_size

bada3a3

Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>

xiaoxiang781216 requested review from pkarashchenko June 8, 2022 02:24

xiaoxiang781216 force-pushed the tls branch from a121fd6 to bada3a3 Compare June 8, 2022 06:08

pkarashchenko approved these changes Jun 8, 2022

View reviewed changes

masayuki2009 approved these changes Jun 8, 2022

View reviewed changes

masayuki2009 merged commit 11c8f31 into apache:master Jun 8, 2022

xiaoxiang781216 deleted the tls branch July 31, 2022 14:21

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

sched/tls: Wrap up_info_size into tls_info_size #6365

sched/tls: Wrap up_info_size into tls_info_size #6365
masayuki2009 merged 2 commits into
apache:masterfrom
xiaoxiang781216:tls

xiaoxiang781216 commented Jun 5, 2022 •

edited

Loading

Uh oh!

masayuki2009 commented Jun 5, 2022

Uh oh!

xiaoxiang781216 commented Jun 5, 2022

Uh oh!

masayuki2009 commented Jun 5, 2022

Uh oh!

xiaoxiang781216 commented Jun 6, 2022

Uh oh!

masayuki2009 commented Jun 6, 2022

Uh oh!

xiaoxiang781216 commented Jun 6, 2022

Uh oh!

xiaoxiang781216 commented Jun 7, 2022 •

edited

Loading

Uh oh!

xiaoxiang781216 commented Jun 7, 2022 •

edited

Loading

Uh oh!

xiaoxiang781216 commented Jun 8, 2022

Uh oh!

masayuki2009 commented Jun 8, 2022

Uh oh!

xiaoxiang781216 commented Jun 8, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

xiaoxiang781216 commented Jun 5, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Impact

Testing

Uh oh!

masayuki2009 commented Jun 5, 2022

Uh oh!

xiaoxiang781216 commented Jun 5, 2022

Uh oh!

masayuki2009 commented Jun 5, 2022

Uh oh!

xiaoxiang781216 commented Jun 6, 2022

Uh oh!

masayuki2009 commented Jun 6, 2022

Uh oh!

xiaoxiang781216 commented Jun 6, 2022

Uh oh!

xiaoxiang781216 commented Jun 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

xiaoxiang781216 commented Jun 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

xiaoxiang781216 commented Jun 8, 2022

Uh oh!

masayuki2009 commented Jun 8, 2022

Uh oh!

xiaoxiang781216 commented Jun 8, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

xiaoxiang781216 commented Jun 5, 2022 •

edited

Loading

xiaoxiang781216 commented Jun 7, 2022 •

edited

Loading

xiaoxiang781216 commented Jun 7, 2022 •

edited

Loading