On 11. 04. 2023. 19:47, Andy Shevchenko wrote:
On Tue, Apr 11, 2023 at 08:15:20PM +0300, Andy Shevchenko wrote:
From: Mirsad Todorovac mirsad.todorovac@alu.unizg.hr
Ah, Mirsad, I need your SoB, if you are okay with the change. Otherwise I have to drop your authorship
Hi, Andy,
Sorry I needed to vent my beer before doing this test:
[root@pc-mtodorov linux_torvalds]# uname -rms Linux 6.2.0-mglru-kmlk-andy-09238-gd2980d8d8265 x86_64 [root@pc-mtodorov linux_torvalds]# tools/testing/selftests/gpio/gpio-sim.sh 1. chip_name and dev_name attributes 1.1. Chip name is communicated to user 1.2. chip_name returns 'none' if the chip is still pending 1.3. Device name is communicated to user 2. Creating and configuring simulated chips 2.1. Default number of lines is 1 2.2. Number of lines can be specified 2.3. Label can be set 2.4. Label can be left empty 2.5. Line names can be configured 2.6. Line config can remain unused if offset is greater than number of lines 2.7. Line configfs directory names are sanitized 2.8. Multiple chips can be created 2.9. Can't modify settings when chip is live 2.10. Can't create line items when chip is live 2.11. Probe errors are propagated to user-space 2.12. Cannot enable a chip without any GPIO banks 2.13. Duplicate chip labels are not allowed 2.14. Lines can be hogged 3. Controlling simulated chips 3.1. Pull can be set over sysfs 3.2. Pull can be read from sysfs 3.3. Incorrect input in sysfs is rejected 3.4. Can't write to value 4. Simulated GPIO chips are functional 4.1. Values can be read from sysfs 4.2. Bias settings work correctly cat: /sys/devices/platform/gpio-sim.0/gpiochip18/sim_gpio0/value: No such file or directory tools/testing/selftests/gpio/gpio-sim.sh: line 393: test: =: unary operator expected bias setting does not work GPIO gpio-sim test FAIL [root@pc-mtodorov linux_torvalds]# git apply ../gpio-sim-andy2.patch [root@pc-mtodorov linux_torvalds]# tools/testing/selftests/gpio/gpio-sim.sh 1. chip_name and dev_name attributes 1.1. Chip name is communicated to user 1.2. chip_name returns 'none' if the chip is still pending 1.3. Device name is communicated to user 2. Creating and configuring simulated chips 2.1. Default number of lines is 1 2.2. Number of lines can be specified 2.3. Label can be set 2.4. Label can be left empty 2.5. Line names can be configured 2.6. Line config can remain unused if offset is greater than number of lines 2.7. Line configfs directory names are sanitized 2.8. Multiple chips can be created 2.9. Can't modify settings when chip is live 2.10. Can't create line items when chip is live 2.11. Probe errors are propagated to user-space 2.12. Cannot enable a chip without any GPIO banks 2.13. Duplicate chip labels are not allowed 2.14. Lines can be hogged 3. Controlling simulated chips 3.1. Pull can be set over sysfs 3.2. Pull can be read from sysfs 3.3. Incorrect input in sysfs is rejected 3.4. Can't write to value 4. Simulated GPIO chips are functional 4.1. Values can be read from sysfs 4.2. Bias settings work correctly GPIO gpio-sim test PASS [root@pc-mtodorov linux_torvalds]#
So the issue was reproduced again on the above mentioned kernel, and the patch fixes it. I will never trust an unexecuted line again, so help me the Lord Above.
Yes, you have both:
Tested-by: Mirsad Goran Todorovac mirsad.todorovac@alu.unizg.hr Signed-off-by: Mirsad Goran Todorovac mirsad.todorovac@alu.unizg.hr
Thank you for considering my bug report. It is great to assist your professional team.
This issue wasn't triggered by the 6.1.15 mglru kernel, and I am wondering what could be the difference? (With the same gpio-sim.sh and patch.)
Best regards, Mirsad