Hello Fabrice,

Fabrice Gasnier [off-list ref] hat am 30. M?rz 2018 um 14:40 geschrieben:


On 09/26/2017 01:59 PM, gohai at sukzessiv.net wrote:

quoted

Notifications for devices without bus or class set get dropped by
dev_uevent_filter. Adding the class to the exported child matches
what the gpio subsystem is doing.

With this change exporting a channel triggers a udev event, which
gives userspace a chance to fixup permissions and makes it possible
for non-root users to make use of the pwm subsystem.

Signed-off-by: Gottfried Haider <redacted>
CC: Thierry Reding <redacted>
CC: H Hartley Sweeten <hsweeten@visionengravers.com>
CC: linux-pwm at vger.kernel.org
CC: linux-arm-kernel at lists.infradead.org
CC: linux-rpi-kernel at lists.infradead.org
---
 drivers/pwm/sysfs.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/pwm/sysfs.c b/drivers/pwm/sysfs.c
index a813239..83f2b0b 100644
--- a/drivers/pwm/sysfs.c
+++ b/drivers/pwm/sysfs.c

@@ -263,6 +263,7 @@ static int pwm_export_child(struct device *parent, struct pwm_device *pwm)
        export->pwm = pwm;
        mutex_init(&export->lock);

+       export->child.class = parent->class;

Hi all,

Sorry to raise this old mail thread. I just figured out this patch is
causing *regression* on v4.16-rcs.

This patch has side effect at my end, with multiple pwm chip. It creates
a new entry in '/sys/class/pwm' every time a 'pwmX' is exported:
- echo X > export

This breaks pwm on platforms that have multiple pwmchip:
- 1st time export will create a /sys/class/pwm/pwmX
- when another export happens on another pwmchip, it can't be created
(e.g. -EEXIST)

I looked at /Documentation/ABI/testing/sysfs-class-pwm:
- pmwX should be there: /sys/class/pwm/pwmchipN/pwmX (only ?)

With this patch:
- pwmX symlink is created in addition, directly under /sys/class/pwm

Example on stm32 (stm32429i-eval) platform:
---
$ ls /sys/class/pwm
pwmchip0 pwmchip4

$ cd /sys/class/pwm/pwmchip0/
$ echo 0 > export

$ ls /sys/class/pwm
pwm0 pwmchip0 pwmchip4

$ cd /sys/class/pwm/pwmchip4/
$ echo 0 > export

sysfs: cannot create duplicate filename '/class/pwm/pwm0'
CPU: 0 PID: 50 Comm: sh Not tainted 4.16.0-rc7-00020-g3361545 #1682
Hardware name: STM32 (Device Tree Support)
[<0000c0f1>] (unwind_backtrace) from [<0000b23b>] (show_stack+0xb/0xc)
[<0000b23b>] (show_stack) from [<0008d2f1>] (sysfs_warn_dup+0x31/0x48)
[<0008d2f1>] (sysfs_warn_dup) from [<0008d4a5>]
(sysfs_do_create_link_sd+0x75/0x88)
[<0008d4a5>] (sysfs_do_create_link_sd) from [<000e8e91>]
(device_add+0x111/0x374)
[<000e8e91>] (device_add) from [<000ca795>] (export_store+0xb5/0x12c)
[<000ca795>] (export_store) from [<0008c899>] (kernfs_fop_write+0x87/0xda)
[<0008c899>] (kernfs_fop_write) from [<0005a0a5>] (__vfs_write+0x1d/0xcc)
[<0005a0a5>] (__vfs_write) from [<0005a1c7>] (vfs_write+0x4f/0x7c)
[<0005a1c7>] (vfs_write) from [<0005a29b>] (SyS_write+0x33/0x70)
[<0005a29b>] (SyS_write) from [<00009001>] (ret_fast_syscall+0x1/0x58)
Exception stack...
-sh: write error: File exists

Not sure what the best fix would be thought :-(

probably pwmX should be named also according with pwmchipN ?
- dev_set_name(&export->child, "pwm%u", pwm->hwpwm);
+ dev_set_name(&export->child, "pwmchip%d-pwm%u", chip->base, pwm->hwpwm);
BUT I think this would break existing ABI...

Also this is quite late in the cycle. Maybe a revert would be wise for now ?

sorry i didn't noticed your mail before. Could you please prepare a revert patch?

Thanks
Stefan