ixgbe is not getting detached on reboot?

ixgbe is not getting detached on reboot?

[Toomas Soome]

[-- Attachment #1: Type: text/plain, Size: 343 bytes --]

hi!

With DEBUG kernel, I’m seeing:

syncing file systems... done
DDI periodic handler not deleted (id=1, hdlr=ixgbe_link_timer)
rebooting…

Since ixgbe detach is calling for cleanup, I guess the message means that ixgbe is otherwise busy and can not be detached? Or, I suspect, something else is going on perhaps?

rgds,
toomas

[-- Attachment #2: Type: text/html, Size: 4889 bytes --]

Re: ixgbe is not getting detached on reboot?

[Andy Fiddaman]

[-- Attachment #1: Type: text/plain, Size: 595 bytes --]

On Thursday, 4 April 2024, at 9:30 AM, Toomas Soome wrote:
> Since ixgbe detach is calling for cleanup, I guess the message means that ixgbe is otherwise busy and can not be detached? Or, I suspect, something else is going on perhaps?


If I'm reading the code correctly, it looks like any time the watchdog does a card reset, you will end up with one extra running timeout as one will be started at the end of ixgbe_reset(), and another back in ixgbe_local_timer() -  N+1 timeouts after N watchdog resets.

We should fix this and switch to ddi_periodic_add() while we're there.

Andy

[-- Attachment #2: Type: text/html, Size: 1302 bytes --]

Re: [developer] ixgbe is not getting detached on reboot?

[Toomas Soome]

[-- Attachment #1: Type: text/plain, Size: 792 bytes --]



> On 4. Apr 2024, at 12:40, Andy Fiddaman <illumos@fiddaman.net> wrote:
> 
> On Thursday, 4 April 2024, at 9:30 AM, Toomas Soome wrote:
>> Since ixgbe detach is calling for cleanup, I guess the message means that ixgbe is otherwise busy and can not be detached? Or, I suspect, something else is going on perhaps?
> If I'm reading the code correctly, it looks like any time the watchdog does a card reset, you will end up with one extra running timeout as one will be started at the end of ixgbe_reset(), and another back in ixgbe_local_timer() -  N+1 timeouts after N watchdog resets.
> 
> We should fix this and switch to ddi_periodic_add() while we're there.
> 

ok, so this rabbit hole is deeper than quick glance did allow to see. I’ll file the issue, thanks!

Toomas


[-- Attachment #2: Type: text/html, Size: 1814 bytes --]