> -----Original Message-----
> From: Jakub Kicinski <[email protected]>
> Sent: Thursday, June 12, 2025 9:22 PM
> To: Haiyang Zhang <[email protected]>
> Cc: [email protected]; [email protected]; Haiyang Zhang
> <[email protected]>; Dexuan Cui <[email protected]>;
> [email protected]; KY Srinivasan <[email protected]>; Paul
> Rosswurm <[email protected]>; [email protected]; [email protected];
> [email protected]; [email protected]; [email protected];
> [email protected]; [email protected]; Long Li <[email protected]>;
> [email protected]; [email protected];
> [email protected]; [email protected]; [email protected];
> [email protected]; [email protected]; [email protected];
> [email protected]; [email protected]; Konstantin
> Taranov <[email protected]>; [email protected]; linux-
> [email protected]
> Subject: [EXTERNAL] Re: [PATCH net-next,v7] net: mana: Add handler for
> hardware servicing events
>
> On Tue, 10 Jun 2025 11:42:22 -0700 Haiyang Zhang wrote:
> > v6:
> > Not acquiring module refcnt as suggested by Paolo Abeni.
>
> TBH I'm not 100% sure this is correct.
> If the service worker operations end up unbinding the driver from
> the device holding the device ref may not prevent the module from
> being unloaded.
>
> Could you try to trigger that condition? Make that msleep() in the work
> even longer and try to remove the module while the work is sleeping
> there?
Thanks for your suggestion! I tested and found that I can rmmod mana
during the sleep and caused accessing freed memory. And getting the extra
module refcnt fixed this (prevented rmmod during sleep). So, I added back
the module refcnt holding, and submitted v8.
Thanks,
- Haiyang
